Progress 08/15/13 to 08/14/20
Outputs
Target Audience:2.1Target Audience Two graduate students in the Department of Mechanical and Aerospace Engineering at UCF were involved in robot experiments and project finalization. 2.2 Impacts or Expected Impacts We published journal papers in the areas of field robots and precision agriculture. We collaborated with several researchers from the University of Florida, the University of California, Merced, and Washington State University in related areas. The project impact is: the developed autonomous cooperative robotic system can reduce the cost of labor in typical disease detection operations. The detection technique can identify stress at an early stage which can benefit many growers. 2.3 Efforts One female graduate student (Madeline Mapes) at UCF has helped the project finalization. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?
Nothing Reported
How have the results been disseminated to communities of interest?Two journal papers were published. What do you plan to do during the next reporting period to accomplish the goals?
Nothing Reported
Impacts
What was accomplished under these goals?
The following tasks were planned for the last year of the project and the percentages of accomplishments are shown along with the tasks. It is worth mentioning that many experiment tasks have not been conducted due to the unprecedented pandemic (laboratory is closed). Ground robot (Goal/Object 1) Further test the motion function of ground robot. (100%) Modify and test the leaf picking mechanism. (0%) Aerial robot (Goal/Object 2): Further enhance the flight control. (0%) Disease detection sensor (Goal/Object 3): Alleviate the operational requirement of the disease detection sensor. (0%) Education, outreach, and dissemination (Goal/Object 4): Submit journal papers and attend conference/meetings. (100%) Supervise >= 2 graduate students. (100%) Support >= 1 postdoc researcher. (0%) Attract and retain students from under-represented groups. (100%) 5.1 Major Activities Conducted to Accomplish Goal/Object 1 The activities relating to the ground scouting robot are: (i) two journal papers have been published in journals; and (ii) the scouting robot was tested in a commercial strawberry field multiple times in testing its scouting capability and the success rate is very high. 5.2 Major Activities Conducted to Accomplish Goal/Object 2: No major activities related to the UAV in the system were carried in this project year. 5.3 Major Activities Conducted to Accomplish Goal/Object 3: UC Merced was responsible for the tasks of this goal/objective. However the subcontract to UC Merced didn't get extended to this project year and there is no expenditure. Therefore, there is no activity relating to goal/objective 3. 5.5 Data collected Experiment related pictures and videos, and design documents are all saved. 5.6 Summary statistics and discussion of results The research product includes a scouting robot, an octorotor, and a disease detection sensor. All of them have been tested. However, due to some practical challenges and unexpected pandemic situations, the integrated system hasn't been successfully validated. Some specific technical challenges we experiences are: (i) the operation requirement of the disease detection sensor is not easy to be satisfied; (ii) the current leaf pick mechanism is slow and its accuracy is largely affected by the vision system. 5.7 Key Outcomes or Other Accomplishments Realized. A change in knowledge: None A change in action: Due to the pandemic situation, many experiment tasks were delayed or cancelled. A change in condition: Due to the unprecedented pandemic situation, some experiment tasks were cancelled.
Publications
- Type:
Journal Articles
Status:
Published
Year Published:
2019
Citation:
P. Menendez-Aponte, X. Kong, and Y. Xu, An Approximated, Control A?ne Model for a Strawberry Field Scouting Robot Considering Wheel Terrain Interaction, Robotica, Vol. 37, Issue 9, Sept. 2019, pp. 1545-1561.
- Type:
Journal Articles
Status:
Published
Year Published:
2020
Citation:
X. Kong, and Y. Xu, Strawberry Plant Alive Status Detection and Relative Pixel Based Plant Localization , ASME Journal of Dynamic Systems, Measurement and Control, May 2020, Vol. 142, pp. 054501-1.
|
Progress 08/15/13 to 08/12/20
Outputs
Target Audience:2.1Target Audience Two graduate students in the Department of Mechanical and Aerospace Engineering at UCF were involved in robot experiments and project finalization. 2.2 Impacts or Expected Impacts We published journal papers in the areas of field robots and precision agriculture. We collaborated with several researchers from the University of Florida, the University of California, Merced, and Washington State University in related areas. The project impact is: the developed autonomous cooperative robotic system can reduce the cost of labor in typical disease detection operations. The detection technique can identify stress at an early stage which can benefit many growers. 2.3 Efforts One female graduate student (Madeline Mapes) at UCF has helped the project finalization. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?
Nothing Reported
How have the results been disseminated to communities of interest?Two journal papers were published. What do you plan to do during the next reporting period to accomplish the goals?
Nothing Reported
Impacts
What was accomplished under these goals?
The following tasks were planned for the last year of the project and the percentages of accomplishments are shown along with the tasks. It is worth mentioning that many experiment tasks have not been conducted due to the unprecedented pandemic (laboratory is closed). Ground robot (Goal/Object 1) Further test the motion function of ground robot. (100%) Modify and test the leaf picking mechanism. (0%) Aerial robot (Goal/Object 2): Further enhance the flight control. (0%) Disease detection sensor (Goal/Object 3): Alleviate the operational requirement of the disease detection sensor. (0%) Education, outreach, and dissemination (Goal/Object 4): Submit journal papers and attend conference/meetings. (100%) Supervise >= 2 graduate students. (100%) Support >= 1 postdoc researcher. (0%) Attract and retain students from under-represented groups. (100%) 5.1 Major Activities Conducted to Accomplish Goal/Object 1 The activities relating to the ground scouting robot are: (i) two journal papers have been published in journals; and (ii) the scouting robot was tested in a commercial strawberry field multiple times in testing its scouting capability and the success rate is very high. 5.2 Major Activities Conducted to Accomplish Goal/Object 2: No major activities related to the UAV in the system were carried in this project year. 5.3 Major Activities Conducted to Accomplish Goal/Object 3: UC Merced was responsible for the tasks of this goal/objective. However the subcontract to UC Merced didn't get extended to this project year and there is no expenditure. Therefore, there is no activity relating to goal/objective 3. 5.5 Data collected Experiment related pictures and videos, and design documents are all saved. 5.6 Summary statistics and discussion of results The research product includes a scouting robot, an octorotor, and a disease detection sensor. All of them have been tested. However, due to some practical challenges and unexpected pandemic situations, the integrated system hasn't been successfully validated. Some specific technical challenges we experiences are: (i) the operation requirement of the disease detection sensor is not easy to be satisfied; (ii) the current leaf pick mechanism is slow and its accuracy is largely affected by the vision system. 5.7 Key Outcomes or Other Accomplishments Realized. A change in knowledge: None A change in action: Due to the pandemic situation, many experiment tasks were delayed or cancelled. A change in condition: Due to the unprecedented pandemic situation, some experiment tasks were cancelled.
Publications
- Type:
Journal Articles
Status:
Published
Year Published:
2019
Citation:
P. Menendez-Aponte, X. Kong, and Y. Xu, An Approximated, Control A?ne Model for a Strawberry Field Scouting Robot Considering Wheel Terrain Interaction, Robotica, Vol. 37, Issue 9, Sept. 2019, pp. 1545-1561.
- Type:
Journal Articles
Status:
Published
Year Published:
2020
Citation:
X. Kong, and Y. Xu, Strawberry Plant Alive Status Detection and Relative Pixel Based Plant Localization , ASME Journal of Dynamic Systems, Measurement and Control, May 2020, Vol. 142, pp. 054501-1.
|
Progress 08/15/18 to 08/14/19
Outputs
Target Audience:2.1 Target Audience In this reporting period, two graduate students in the Department of Mechanical and Aerospace Engineering at UCF were involved in robot experiments and a new flight control algorithm design for the UAV. One graduate student and one postdoc researcher majoring in Mechanical Engineering and Electrical Engineering were actively involved in the disease detection sensor research at UC Merced. The major target audience for this project includes strawberry growers, crop consultants, crop scientists, and horticulturists. The research progress has been demonstrated in the following events: 2018 ASME Dynamic Systems and Controls Conference, September 30-October 3, 2018, Atlanta, Georgia, USA. IX North American Strawberry Symposium, Feb. 2019, Orlando, Florida. 2019 American Society for Horticultural Science (ASHS) Annual Conference, July 2019, Las Vegas, Nevada. Field experiments of the ground disease detection robot continued in Pappy's Patch, Oviedo FL. The functionalitytests were still focusing on the scouting control. We tested a new plant detection and localization algorithm. Also we are in the progress of developing a new reinforcement learning based controller for the octorotor. 2.2. Impacts or Expected Impacts In this reporting period, we have the following impacts: We submitted and published conference and journal papers in the areas of field robots and precision agriculture. We collaborated with several researchers in related areas. The project impact is: the autonomous cooperative robotic system has the potential of dramatically reducing labor costs in disease detection operations. Currently the disease detection process is manual, time consuming, and can be late in detecting diseases. The disease and stress detection technique that can identify the stress at an early stage is of great interest to many growers. 2.3 Efforts In the reporting period, one female graduate student at UCF has helped the robotdesign and testing: Xiangling Kong. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?N/A How have the results been disseminated to communities of interest?We present papers/posters in the following events: 2018 ASME Dynamic Systems and Controls Conference (September 30-October 3, 2018, Atlanta, Georgia), the IX North American Strawberry Symposium (Feb. 2018, Orlando, Florida), and 2019 American Society for Horticultural Science (ASHS) Annual Conference (July 2019, Las Vegas, Nevada). We published 4 journal and 3 conference papers. We have contributed 2 posters. What do you plan to do during the next reporting period to accomplish the goals?The following tasks are planned for Year 7 (August 15, 2019 - August 14, 2020). Ground robot: Further testthe motion function of the ground robot Modify the leaf picking mechanism Strawberry aerial robot: Further enhance the flight control Disease Detection Sensor: Alleviate the operational requirement of the disease detection sensor Education, Outreach, and Dissemination: Submit journal papers and attend conference/meetings. Supervise >= 1 graduate student
Impacts
What was accomplished under these goals?
Potential Impact: As described in the previous years' reports, early and efficient disease infection management is a critical task in fruit (e.g. strawberry and citrus) productions as it affects fruit quality, yield, and ultimately market value. Costly and labor intensive manual scouting is currently the most frequently used method of disease and stress detection in crops. If the networked aerial and ground robot system equipped with the new disease detection sensor is successfully developed, commercialized, and adopted by producers/growers, it could potentially reduce their scouting costs by 50% and improve the accuracy of stress detection by 30%, which could further increase their profits. Additionally, if the project is successful, interests in precision farming and agriculture automation using robotic technologies are expected. Project Goals and Objectives: The proposed research tasks leverage strong experiences in cooperative robots' decision making at UCF and fruit disease detection research conducted at UF (now at UC Merced) to reach the ultimate goal of reducing scouting costs and crop loss due to several different stress factors. Specific Goals: A robust aerial crop health monitoring system is in the process of generating geo-referenced crop health maps and high level fruit-specific formation configurations (year 1 to year 3). These maps are then to be supplied to the cooperative agricultural ground platform (year 1 to year 5) equipped with novel sensing tools to navigate to the aerially-diagnosed plant/tree locations and perform detailed sensing tasks. The integrated system will then be evaluated extensively in commercial farms under realistic field conditions followed by overall modification to the network system based on previous findings (year 2 to year 5). The enhanced system will then be tested and validated in commercial farms (year 4 to year 7). Year 6 Project Goals and Milestones: The following tasks were planned for Year 6 (August 15, 2018 - August 14, 2019). The percentages of accomplishments are shown along with the tasks. Ground robot (Goal/Object 1) Integrate the leaf pick algorithm into the main robot software (100%). Test the functionalities of fully integrated robot (cross-bed motion, over-bed motion, disease detection, and leaf picking) (100%). Aerial robot (Goal/Object 2): Conduct more flight tests and develop the communication between the ground robot and the UAV. (50%) Disease detection sensor (Goal/Object 3): Developenew hardware with a wider field of view and less light requirement. Design a new algorithm for training and classification of exposed samples. Field testthe system and comparethe accuracy against a spectrophotometer. Education, outreach, and dissemination (Goal/Object 4): Submit journal papers and attend conference/meetings. (100%) Supervise >= 2 graduate students (100%) Support >= 1 postdoc researcher (100%) Attract and retain students from under-represented groups. (100%) Major Activities Conducted to Accomplish Goal/Object 1 The research conducted in the area of ground scouting robot is summarized as follows. We further tested our scouting functions in the Pappy's Patch for more than three times over the strawberry season. With onboard range finders, RGB cameras, and controllers, the robot is able to autonomously move in, along, and out of strawberry beds for multiple rows. The results have been published in the ASME Journal of Dynamic Systems, Measurement and Control. We are investigating a new color detection method using RGB cameras and the salient feature of the method is its robustness with respect to varying light scenes. With this low-cost method, we are able to conduct strawberry plant localization, achieving a center-meter level accuracy. Major Activities Conducted to Accomplish Goal/Object 2: The focus of the octorotor research this year is on a reinforcement learning based flight control. This type of control is able to achieve an optimal control performance, while dealing with uncertainty issues in octorotor dynamics and sensor/actuator noise. The drawback of existing reinforcement learning controllers is thehigh computational cost. To tackle this issue, we investigated search space dimension reduction methods. The simulation results show that the optimal control commands can be generated in real-time. Major Activities Conducted to Accomplish Goal/Object 3: Enhancement was done on developing a low-cost disease detection camera. The system includes two cameras (RGB and NIR) with small lenses connected to a single board computer with an interface shield. The system has the ability to be controlled remotely and transfers collected data via Wi-Fi communication. This sensor can be installed on the field robotor the UAV. After the initial experimentin the lab, we set up and installed the system for field data collection. Several strawberry plants were sampled randomly for this purpose. Based on the collected data, the images were processed and analyzed with the developed algorithm in Python. The developed algorithm was able to collect the output data fromtwo CCD cameras in both visible and near infrared bands. The two image data matrices overlaid and matched with respect to the distance between each sensor to the object and their field of view. The output matrix consists of the calculated NDVI for each pixel of the corrected field of view. The software system developed for this system were able to calculate different vegetation indices (NDVI) using both RGB and NIR bands. Our comparative study showed that the developed vision sensor can provide and generate accurate results which are in an agreement with the extracted data from the HR-1024 spectrometer. Data Collected We have collected and saved experimental pictures and videos, design documents, software, publications, and presentation. Summary Statistics and Discussion of Results The whole system consists of octorotor UAV, ground robot, and disease detection sensors. All of them have been experimentally tested. However, the connections and communications between the UAV and the ground robot haven'tbeen accomplished. Also there are some practical challenges involved in integrating the disease detection sensor with the ground robot: the operation requirement of the disease detection sensor is not easy to be satisfied. We will address these issues in the next reporting period. Key Outcomes or Other Accomplishments Realized. A change in knowledge: There is no knowledge change between the reports of last year and this year. The knowledge we gained in this reporting period has been or will be presented in journal papers, conferences, and meetings. A change in action: Same as last year, mainly graduate students and postdoc researcherare involved in the research. We mainly worked on programming, integration, and testing tasks. A change in condition: There is no change in conditions as compared with last year.
Publications
- Type:
Other
Status:
Published
Year Published:
2018
Citation:
Poster: Ehsani R. 2018. Sensor System for Monitoring Horticultural Crops: Challenges and Opportunities. Phenome 2018 conference, Tucson, AZ
- Type:
Other
Status:
Published
Year Published:
2018
Citation:
Poster/Presentation: Ehsani R. 2018. AgTech Research in High value crops. UC Merced AgTech day, Merced, CA.
- Type:
Theses/Dissertations
Status:
Published
Year Published:
2019
Citation:
Xiangling Kong Color-Ratio Based Strawberry Plant Localization and Nutrient Deficiency Detection , August 2019. PhD Dissertation, UCF
- Type:
Journal Articles
Status:
Published
Year Published:
2019
Citation:
Pablo Menendez-Aponte, Xiangling Kong, Yunjun Xu, An Approximated, Control Affine Model for A Strawberry Field Scouting Robot Considering Wheel Terrain Interaction, Robotica, March 05, 2019
- Type:
Journal Articles
Status:
Published
Year Published:
2019
Citation:
Douglas Freese, and Yunjun Xu, Multi-Phase Scouting Control of an Agricultural Field Robot with Reachability Analyses, the Journal of Dynamic Systems, Measurement and Control, 141(5), 051009 (Jan 29, 2019),
- Type:
Journal Articles
Status:
Published
Year Published:
2018
Citation:
She, Y., R. Ehsani, J. Robbins, J. Nah�n Leiva, and J. Owen. 2018. Applications of High-Resolution Imaging for Open Field Container Nursery Counting. Remote Sensing. 10 (12).
- Type:
Journal Articles
Status:
Published
Year Published:
2018
Citation:
Wan, P., A. Toudeshki, H. Tan, and R. Ehsani. 2018. A methodology for fresh tomato maturity detection using computer vision. Computers and Electronics in Agriculture, 146, 43-50.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2018
Citation:
S. Defterli, Y. Xu, Virtual Camouflage Based Visual Servo Control of a Leaf Picking Mechanism, the 2018 ASME Dynamic Systems and Controls Conference, September 30 - October 3, 2018, Atlanta, Georgia, USA
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2018
Citation:
X. Kong, Y. Xu, Strawberry Plant Localization via Relative Pixels in Sequential Images, the 2018 ASME Dynamic Systems and Controls Conference, September 30-October 3, 2018, Atlanta, Georgia, USA
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2018
Citation:
Q. Li, Y. Xu, Image Quality-Driven Octorotor Flight Control via Reinforcement Learning, the 2018 ASME Dynamic Systems and Controls Conference, September 30-October 3, 2018, Atlanta, Georgia, USA
|
Progress 08/15/17 to 08/14/18
Outputs
Target Audience:Target Audience In this reporting period, four graduate students in the Department of Mechanical and Aerospace Engineering at UCF were involved in the robot enhancement, testing, and new software development and upgrade. One graduate student and one postdoc researcher at UC Merced were involved in the software design and upgrade for the spectrometer. The research progress has been demonstrated in the following events: The 5th NRI PI meeting (November 2017, Washington DC). 2018 ASABE Annual International Meeting, July 2018, Detroit, Michigan. 2017 ASME Dynamic Systems and Control Conference (October 2017, Virginia). Phenome 2018, Tucson Arizona, February 14-17, 2018. UC Merced Ag Tech Day Conference, May, 4th, 2018. Agrologic Food Tech Summit, San Francisco, July 18-19, 2018. The ground disease detection robot has been extensively tested in Pappy's Patch, Oviedo FL. The functionally tests were focusing on scouting control, wheel terrain interaction model, manipulator/arm control, and leaf detection. The UC Merced team present the research progress at three different conferences. The total number of participants in these events is more than 600. The audience include scientists, growers, and representatives from food processors and commodity groups. Impacts or Expected Impacts In this reporting period, we have the following impacts: Conference and journal papers in the areas of field robots and precision agriculture are accepted or published. We have initiated some collaborations with several researchers. The project impact is: the autonomous cooperative robotic system has the potential of dramatically reducing labor costs in disease detection operations. Currently the disease detection process is manual, time consuming, and can be late in detecting diseases. The disease and stress detection technique that can identify the stress at an early stage is of great interest to many growers. Efforts In the reporting period, two female graduate students at UCF and one female student at UC Merced have helped the robot and disease detector designs: Sinem Defterli, Xiangling Kong, and Elnaz Akbari. Changes/Problems:Key Outcomes or Other Accomplishments Realized. A change in knowledge: There is no knowledge change between the reports of last year and this year. The knowledge we gained in this reporting period has been or will be presented in journal papers, conferences, and meetings. A change in action: Same as last year, mainly graduate students and postdoc researchers are involved in the research. Programming, integration, and testing are mainly the tasks we worked on. A change in condition: There is no change in conditions as compared with last year. What opportunities for training and professional development has the project provided?No How have the results been disseminated to communities of interest?We present papers/posters in the following events: the 5th NRI PI meeting (November 2017, Washington DC), 2018 ASABE Annual International Meeting (July 2018, Detroit, Michigan), 2017 ASME Dynamic Systems and Control Conference (October 2017, Virginia), Phenome 2018 (Tucson Arizona, February 14-17, 2018), UC Merced Ag Tech Day Conference (May, 4th, 2018), and Agrologic Food Tech Summit (San Francisco, July 18-19, 2018). We published 1 journal and 1 conference papers. We did 7 presentations and contributed to 3 posters. Also we have 3 conference papers being accepted and currently two journal papers are under review. What do you plan to do during the next reporting period to accomplish the goals?The following tasks are planned for Year 6 (August 15, 2018 - August 14, 2019). Ground robot: Integrate the leaf pick algorithm into the main robot software. Test the fully integrated functionality of the robot (cross-bed motion, over-bed motion, disease detection, and leaf picking). Strawberry aerial robot: Conduct more flight tests and develop the communication between the ground robot and the UAV. Disease Detection Sensor: Test the disease detection functionality onboard the ground robot in a commercial farm. Education, Outreach, and Dissemination: Submit journal papers and attend conference/meetings. Supervise >= 2 graduate students Support >= 1 postdoc researcher. Attract and retain students from under-represented groups.
Impacts
What was accomplished under these goals?
Potential Impact: As described in the previous years' reports, early and efficient disease infection management is a critical task in fruit (e.g. strawberry and citrus) productions as it affects fruit quality, yield, and ultimately market value. Costly and labor intensive manual scouting is currently the most frequently used method of disease and stress detection in crops. If the networked aerial and ground robot system equipped with the new disease detection sensor is successfully developed, commercialized, and adopted by producers/growers, it could potentially reduce their scouting costs by 50% and improve the accuracy of stress detection by 30%, which could further increase their profits. Additionally, if the project is successful, interests in precision farming and agriculture automation using robotic technologies are expected. Project Goals and Objectives: The proposed research tasks leverage strong experiences in cooperative robots' decision making at UCF and fruit disease detection research conducted at UF (now at UC Merced) to reach the ultimate goal of reducing scouting costs and crop loss due to several different stress factors. Specific Goals: A robust aerial crop health monitoring system is in the process of generating geo-referenced crop health maps and high level fruit-specific formation configurations (year 1 to year 3). These maps are then to be supplied to the cooperative agricultural ground platform (year 1 to year 5) equipped with novel sensing tools to navigate to the aerially-diagnosed plant/tree locations and perform detailed sensing tasks. The integrated system will then be evaluated extensively in commercial farms under realistic field conditions followed by overall modification to the network system based on previous findings (year 2 to year 5). The enhanced system will then be tested and validated in commercial farms (year 4 to year 6). Year 5 Project Goals and Milestones: The following tasks were planned for Year 5 (August 15, 2017 - August 14, 2018). The percentages of accomplishments are shown along with the tasks. Ground robot (Goal/Object 1) Leaf sampling control system design will be finalized (95%). Test the field scouting functionality of the robot (100%) Test the fully integrated robot (cross-bed motion, over-bed motion, disease detection, and leaf picking) (80%) Aerial robot (Goal/Object 2): Real-time geo-location functionaltiy via flight test (90%). Disease detection sensor (Goal/Object 3): Integrate the disease detection sensor with the ground robot (50%). Test the disease detection functionality onboard of the ground robot in a commercial farm (10%). Note: Objective 3 is not achieved mainly because the funding transfer from UF to UC Merced was delayed for about 8 months. Education, outreach, and dissemination (Goal/Object 4): Submit journal papers and attend conference/meetings. (100%) Supervise >= 4 graduate students (100%) Support >= 1 postdoc researcher. (100%) Attract and retain students from under-represented groups. (100%) Reach out to one more growers/producers who are potentially interested in this research product. (0%) 5.1 Major Activities Conducted to Accomplish Goal/Object 1 The main research activities on the ground robotare leaf sampling control and scouting control. Firstly, in terms of the leaf sampling control, a virtual motion camouflage based visual servoing algorithm has been developed for the picking mechanism to reach the target leaf in real-time and with minimized energy consumption. The leaf mechanism used the camera-in-hand configuration. The final error between the target and actual end-effector positions has been significantly decreased. The results will be presented in the 2018 ASME DSCC in Atlanta, GA. Secondly, our team has been successfully tested the robot scouting control in a commercial field. The robot can autonomously move over beds and cross beds without scratching bed plastic covers. In the over-bed motion, range finders are used; while in the cross-bed motion, vision sensors are used. A proportional-integral-derivative controller is developed for the over-bed motion, and a vision based sliding mode control is designed for the cross-bed motion. As shown in our submitted journal paper (the Journal of Dynamic Systems, Measurement and Control), the robot successfully scouts over two long strawberry rows autonomously with only a small final heading angle error. In addition to the above two main tasks, we have revisited the wheel-terrain interaction model and investigated a new method of plant geo-referencing via relative pixel information. A different approximation functions were proposed to curve fit the wheel-terrain interaction forces, and unknown parameters are estimated using a Kalman filter in real-time. To enhance the geo-referencing accuracy, a Kalman filter is designed based on relative pixel locations between adjacent plants, and the accuracy is around 1 center meter. This approach is low cost and light weight as compared with a typical approach of using RTK-GPS, and the accuracy is at least the same. Also we are developing a new filtering method to detect leaf color, and the method is invariant to light conditions and relative position/attitude between cameras and leafs. 5.2 Major Activities Conducted to Accomplish Goal/Object 2: For the UAV research, we focused on the development of an image quality driven flight control system. The UAV technologies for information acquisition and rapid responses rely on the quality of the images obtained and the way they are processed. Aerial image quality can be blurry and of a low resolution due to unsteady aircraft motion and non-ideal flight status. We modeled the relationships among image blur rate, image resolution, flight altitude, and flight speed in the horizontal direction. These relationships lead to control commands for the altitude and velocity controllers designed using a reinforcement learning algorithm. Both computer vision algorithms and designed controllers were proven to be successful through experiments and simulations. The result will be presented in the 2018 ASME DSCC in Atlanta, GA. 5.3 Major Activities Conducted to Accomplish Goal/Object 3: We continued the disease detection sensor design. We developed the imaging/spectral analysis algorithm and software. The detection success rate should reach 60%, and the imaging processing speed should reach 0.1Hz (10 seconds processing time). 5.5 Data collected Design documents, hardware/software, testing data (images and videos), publications, presentations, and outreach and dissemination activity pictures are collected and saved. 5.6 Summary statistics and discussion of results The whole system includes three parts: ground robot, octorotor, and disease detection sensors. Currently almost all the subsystems are ready (except a few subsystems such as the communication between the UAV and the ground robot). Mostof the integration tasks are accomplished (except the final testing of the spectrometer). Next year, we will focus on the last few steps in integration and will test the system in the beginning of strawberry season.
Publications
- Type:
Journal Articles
Status:
Published
Year Published:
2017
Citation:
S. G. Defterli, Y. Xu, Analytical Solution of a Five-Degree-of-Freedom Inverse Kinematics Problem for the Handling Mechanism of an Agricultural Robot, International Journal of Mechanisms and Robotic Systems, Vol. 4, Issue 1, 2017.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2017
Citation:
D. Freese, Y. Xu, Nonlinear Robust path Control for a field Robot Scouting in Strawberry orchards, Proceedings of the ASME Dynamic Systems and Controls Conference, October 11-13, 2017, Tysons Corner, Virginia, USA.
- Type:
Conference Papers and Presentations
Status:
Other
Year Published:
2018
Citation:
Arash, R. Ehsani, S. G. Defterli, and Y. Xu, Development of A low-cost Spectrometer Sensor for a Disease Detection Robot, 2018 ASABE Annual International Meeting, Detroit Michigan, July 29 - August 1, 2018.
- Type:
Conference Papers and Presentations
Status:
Other
Year Published:
2018
Citation:
S. G. Defterli and Y. Xu, "Virtual Motion Camouflage based Visual Servo Control of a Leaf Picking Mechanism" 2018 ASABE Annual International Meeting, Detroit Michigan, July 29 - August 1, 2018.
- Type:
Conference Papers and Presentations
Status:
Other
Year Published:
2018
Citation:
Q. Li and Y. Xu, Real-time and Onboard Leaf Position Estimation using an Octo-rotor Controlled by Image Qualities 2018 ASABE Annual International Meeting, Detroit Michigan, July 29 - August 1, 2018.
- Type:
Conference Papers and Presentations
Status:
Other
Year Published:
2018
Citation:
X. Kong and Y. Xu, Strawberry Plant Localization via Relative Pixel Positions in Sequential Images 2018 ASABE Annual International Meeting, Detroit Michigan, July 29 - August 1, 2018.
- Type:
Other
Status:
Other
Year Published:
2018
Citation:
Ehsani, R. Sensor System for Monitoring Horticultural Crops: Challenges and Opportunities. Phenome 2018, Tucson Arizona, February 14-17, 2018.
- Type:
Other
Status:
Other
Year Published:
2018
Citation:
Ehsani, R. AgTech Research in High Value Crops. Ag Tech Day. UC Merced Ag Tech Day conference. May, 4th, 2018.
- Type:
Other
Status:
Other
Year Published:
2018
Citation:
Ehsani, R. The Power of Digital Observation in Agriculture. Agroliogic Food Tech Summit. San Francisco. July 18-19, 2018.
- Type:
Other
Status:
Other
Year Published:
2017
Citation:
The 5th NRI PI meeting (November 2017, Washington DC)
- Type:
Other
Status:
Other
Year Published:
2018
Citation:
Toudeshki A., R. Ehsani, S. G. Defterli, Y. Xu. 2018. Development of A low-cost Spectrometer Sensor for a Disease Detection Robot, 2018 ASABE Annual International Meeting, 76-1800969.
- Type:
Other
Status:
Other
Year Published:
2018
Citation:
Akbari E., A.Toudeshki, A. Alizadeh, R. Ehsani, Y. Xu. 2018. Low-Cost NDVI Camera for Application of Robotic Disease Detection, 2018 ASABE Annual International Meeting, 87-1800590
- Type:
Conference Papers and Presentations
Status:
Accepted
Year Published:
2018
Citation:
X. Kong, Y. Xu, Strawberry Plant Localization via Relative Pixels in Sequential Images, Submitted to the 2018 ASME Dynamic Systems and Controls Conference, September 30-October 3, 2018, Atlanta, Georgia, USA.
- Type:
Conference Papers and Presentations
Status:
Accepted
Year Published:
2018
Citation:
S. Defterli, Y. Xu, Bio-Inspired Visual Servoing Algorithm in Leaf Picking, Submitted to the 2018 ASME Dynamic Systems and Controls Conference, September 30 - October 3, 2018, Atlanta, Georgia, USA.
- Type:
Conference Papers and Presentations
Status:
Accepted
Year Published:
2018
Citation:
Q. Li, Y. Xu, Image Quality-Driven Octorotor Flight Control via Reinforcement Learning, Submitted to the 2018 ASME Dynamic Systems and Controls Conference, September 30-October 3, 2018, Atlanta, Georgia, USA.
- Type:
Journal Articles
Status:
Under Review
Year Published:
2018
Citation:
Douglas Freese, and Yunjun Xu, Vision Based Nonlinear Robust Scouting Control in Agricultural Fields with Reachability Analyses, Submitted to the Journal of Dynamic Systems, Measurement and Control, Under Review.
- Type:
Journal Articles
Status:
Under Review
Year Published:
2018
Citation:
Pablo Menendez-Aponte, Xiangling Kong, Yunjun Xu, An Approximated, Control A?ne Model for A Strawberry Field Scouting Robot Considering Wheel Terrain Interaction, Robotica, revision submitted.
- Type:
Theses/Dissertations
Status:
Published
Year Published:
2018
Citation:
Sinem Defterli Bio-inspired Visual Servo Control of a Picking Mechanism in an Agriculture Ground Robot , July 2018.
- Type:
Theses/Dissertations
Status:
Published
Year Published:
2018
Citation:
Douglas Freese Navigation of an Autonomous Differential Drive Robot for Field Scouting in Semi-Structured Environments , April 2018.
|
Progress 08/15/16 to 08/14/17
Outputs
Target Audience:Target Audience In this reporting period, 5 graduate students and 1 postdoc researcher majoring in the mechanical and aerospace engineering and the agriculture engineering were actively involved in the robot design modification, testing, and new software development and upgrade for automation. They were involved in the testing of robots (ground and aerial) at UCF and the disease detection sensor software at UF. The research progress has been demonstrated in the following events: The 4th NRI PI meeting (November 2016, Washington DC) 2016 ASABE International conference (July 2017, Spokane WA) The 5th IFAC Conference on Sensing, Control and Automation for Agriculture, August 14-17, 2016, Seattle, WA. The 7th International Workshop on Collaborative Robots and Human Robot Interaction, November 2016, Orlando, FL. The 8th IGRSM International Conference and Exhibition on Geospatial & Remote Sensing, Kuala Lumpur, Malaysia, April 13-14, 2016 (Not reported in last year's annual report). UAS Technology for Precision Agriculture Application, A workshop at ICUAS 2016, Washington D.C. June 7, 2016(Not reported in last year's annual report). We have continued our extensive robot control testing in Pappy's Patch, Oviedo FL. Impacts or Expected Impacts In this reporting period, we have the following impacts: We have more than 5 professors from different universities in US toured the UCF lab during their seminar visits, and many of them showed great interests in this project. We have submitted conference and journal papers in the areas of robotics and precision farming, some of which were accepted and published. The project impact is: the autonomous cooperative robotic system has the potential of dramatically reducing labor cost in disease detection operations. Currently the disease detection process is manual, time consuming, and can be late in detecting diseases. The disease and stress detection technique that can identify the stress at an early stage is of great interest to many growers. Efforts In the reporting period, one graduate student from UCF is Hispanic. Also two female graduate students at UCF have helped the robot and disease detector designs: Sinem Defterli and Xiangling Kong. Changes/Problems:N/A What opportunities for training and professional development has the project provided?Professional training and development activities were not planned in this reporting period. How have the results been disseminated to communities of interest?Three approaches are used to disseminate the research results. We present papers/posters in the 4th NRI PI meeting, 2017 ASABE international conference, the 5th IFAC Conference on Sensing, Control and Automation for Agriculture, and the 7th International Workshop on Collaborative Robots and Human Robot Interaction, the 8th IGRSM International Conference and Exhibition on Geospatial & Remote Sensing, Kuala Lumpur, Malaysia, April 13-14, 2016, and the UAS Technology for Precision Agriculture Application, A workshop at ICUAS 2016, Washington D.C. June 7, 2016 (Not reported in last year's annual report). 2. We published 2 journals and 2 conference papers, made 4 conference presentations and present 1 poster. Also we have 1 journal and 1 conference papers accepted. 3. We continue our demo and experiment tests in commercial farms, such as the Pappy's Patch to attract the attention from general public. What do you plan to do during the next reporting period to accomplish the goals?The following tasks are planned for Year 5 (August 15, 2017 - August 14, 2018). Ground robot: Leaf sampling control system design will be finalized. Test the fully integrated field navigation functionality of the robot (cross-bed motion, over-bed motion, disease detection, and leaf picking). Strawberry aerial robot: Real-time geo-location functionality via flight test. Disease Detection Sensor: Integrate the disease detection sensor with the ground robot. Test the disease detection functionality onboard the ground robot in a commercial farm. Education, Outreach, and Dissemination: Submit journal papers and attend conference/meetings. Supervise >= 4 graduate students Support >= 1 postdoc researcher. Attract and retain students from under-represented groups. Reach out to one more growers/producers who are potentially interested in this research product.
Impacts
What was accomplished under these goals?
Year 4 Project Goals and Milestones: The following tasks were planned for Year 4 (August 15, 2016 - August 14, 2017). The percentages of accomplished tasks are shown. Ground robot (Goal/Object 1) Finish the development of cross-the-bed control algorithm and test it in at least one commercial farm. (100%) Test the fully integrated field navigation functionality of the robot. (70%) New path/trajectory planning algorithm will be designed and tested based on the identified robot dynamic model. (30%) Leaf sampling control system design will be finalized. (70%) Strawberry aerial robot (Goal/Object 2): Real-time geo-location functionality via flight test. (80%) Disease detection sensor (Goal/Object 3): Continue the disease detection sensor design (100%) Develop the imaging/spectral analysis algorithm and software. The detection success rate should reach 80%, and the imaging processing speed should reach 0.1Hz (10 seconds processing time) (100%) Education, outreach, and dissemination (Goal/Object 4): Submit journal papers and attend conference/meetings. (100%) Supervise >= 4 graduate students (100%) Support >= 1 postdoc researcher. (100%) Attract and retain students from under-represented groups. (100%) Reach out to one more growers/producers who are potentially interested in this research product. (100%) Conduct a first round cost analysis of the autonomous ground robot in disease detection scouting. (0%) 5.1 Major Activities Conducted to Accomplish Goal/Object 1 A new point detection algorithm was developed and tested with samples from multiple fields. This method proved to be more robust to the lighting condition and have a higher accuracy than the gradient based technique (e.g. the Hough Transform method). Although the required processing time is higher, it is still at an acceptable frequency to work within the control loop. The over-the-bed algorithm was enhanced to the PID based controller and included filtering to reject inaccurate readings. The overall loop time was also modified to function with all the other data stream rates and the cross-bed data information. At a moderate speed of 15 cm/sec, the over-the-bed control performed both forwards and in reverse, transferring the vehicle safely over the strawberry row for the desired travel distance of up to 70 meters. The numerical and laboratory simulation part of the cross-bed functionality was conducted to demonstrate the capability of the nonlinear controllers used. The cross-bed controller has been tested and verified in field experiments. The overall accuracy was sufficient based on the final orientation of the vehicle with respect to the strawberry row. From the initial set-up for the vehicle, an offset to the left was about 25 cm, by travelling perpendicular to the row this was reduced to about 3 cm. For the second control phase (performed after rotating to face the row) the initial conditions were a heading offset of about 10 degrees to the left of the row and about 42 cm farther away than desired. The controller converged to be about 1 degree off from the heading and about 2 cm shy of the desired location towards the bed. Integration of the control phases into a single combined test has been successfully performed in simulation. The testing of the full transfer from one row to another will be commenced as soon as the strawberry season has returned. 5.2 Major Activities Conducted to Accomplish Goal/Object 2: There is mainly one achievement in this reporting period: implementation of geo-referencing functionality in real-time and onboard the octo-rotor. Last year, we developed a data-association algorithm to differentiate multiple diseased leafs and used an EKF to find global coordinates for their corresponding measurements. However, all these are based on post-processing the image data recorded during the flight of octo-rotor. To achieve a real-time and onboard implementation, software architecture and several hardware are modified. In terms of software architecture, receiver, IMU, GPS, and barometer software are hosted in an APM flight controller, while all the other software are running in a Raspberry-PI using Python, including image processing for feature detection and target identification, and target global coordinate calculation for 3D position using discrete Kalman filter. A memory is added in Raspberry PI so we can store all the software and required operating system onboard. Currently the working frequency of overall system including camera, transmitter, APM flight controller, Raspberry PI, and associated software is about 6Hz. 5.3 Major Activities Conducted to Accomplish Goal/Object 3: Develop the imaging/spectral analysis algorithm and software. The detection success rate reaches 60%, and the imaging processing speed reaches 0.1Hz (10 seconds processing time). 5.5 Data collected All the design documents, testing data, and outreach and dissemination activity pictures are collected and saved. 5.6 Summary statistics and discussion of results The whole system includes three parts: ground robot, octorotor, and disease detection sensors. There are many hardware and software subsystems involved. After four years of research and development, majority of the subsystems are functioning properly. Next year, we will focus on integration of sensors onto the ground robot, finalizing vision servo based leaf picking mechanism, and extensive field testing. 5.7 Key Outcomes or Other Accomplishments Realized. A change in knowledge: There is no knowledge change between the reports of last year and this year. The key outcome will be the fully autonomous robot that can efficiently scout a commercial strawberry orchard considering scalability/adaptability, lighting control, low cost, and ability to transverse over sandy/grass terrains. Also the new detection sensor will be able to detect water stress in its early stage, which will be onboard of the ground robot and aerial platform. In this reporting period, the knowledge we gained has been presented in journal papers, conferences, meetings, and news. A change in action: Same as last year, mainly graduate students and postdoc researchers are involved in the research. Programming, integration, and testing are mainly the tasks we worked on. A change in condition: There is no change in conditions as compared with last year.
Publications
- Type:
Journal Articles
Status:
Published
Year Published:
2017
Citation:
Jinzhu Lu, Reza Ehsani, Yeyin Shi, Jaafar Abdulridha, Ana I. de Castro, Yunjun Xu, Field detection of anthracnose crown rot in strawberry using spectroscopy technology , Computers and Electronics in Agriculture, 135, 2017, pp. 289-299.
- Type:
Journal Articles
Status:
Published
Year Published:
2017
Citation:
C. Garcia, Y. Xu, Multi-Target Geolocation via an Agricultural Octorotor based on Orthographic Projection and Data Association, the Journal of Aerospace Engineering, Proceedings of iMeche, Part G, Published online on May 31, 2017.
- Type:
Journal Articles
Status:
Accepted
Year Published:
2017
Citation:
S. G. Defterli, Y. Xu, Analytical Solution of a Five-Degree-of-Freedom Inverse Kinematics Problem for the Handling Mechanism of an Agricultural Robot, International Journal of Mechanisms and Robotic Systems, Accepted.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2016
Citation:
C. Garcia, Y. Xu, Target Geolocation for Disease Detection Applications via an Octorotor, the 5th IFAC Conference on Sensing, Control and Automation for Agriculture, August 14-17, 2016, Seattle, Washington, USA.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2016
Citation:
Pablo, M., Garcia, C. A., Freese, D., Defteli, S., and Xu, Y., Software and Hardware Architectures in Cooperative Aerial and Ground Robots for Agricultural Disease Detection, the 7th International Workshop on Collaborative Robots and Human Robot Interaction, Orlando, FL, 2016.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2017
Citation:
Freese, Douglas, and Xu, Yunjun, Sliding Mode Control for an Agricultural Robot with Vision Feedback, 2017 ASABE Annual International Meeting, July 16-19, 2017, Spokane, Washington USA
- Type:
Other
Status:
Other
Year Published:
2016
Citation:
The 4nd NRI PI meeting (Nov. 2016, Washington DC).
- Type:
Theses/Dissertations
Status:
Published
Year Published:
2016
Citation:
Pablo Menendez-Aponte: Development of a control affine model for an agricultural field robot, Dec. 2016
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2017
Citation:
Defterli, Sinem, G., and Xu, Yunjun, Image Based Visual Servo Control of a 6-DoF Leaf Picking Mechanism, 2017 ASABE Annual International Meeting, July 16-19, 2017, Spokane, Washington USA
- Type:
Other
Status:
Other
Year Published:
2016
Citation:
Ehsani, R. The rise of small UAVs: applications, opportunities, and challenges. 8th IGRSM International Conference and Exhibition on Geospatial & Remote Sensing. Kuala Lumpur, Malaysia. 13-14 April, 2016. (Not reported in last years annual report).
- Type:
Other
Status:
Other
Year Published:
2016
Citation:
Ehsani, R. UAS Technology for Precision Agriculture Applications (AgDroneTech16). A workshop at ICUAS 2016, Washington D.C. June 7, 2016. (Not reported in last years annual report).
- Type:
Conference Papers and Presentations
Status:
Accepted
Year Published:
2017
Citation:
D. Freese, Y. Xu, Nonlinear Robust path Control for a field Robot Scouting in Strawberry orchards, Proceedings of the ASME Dynamic Systems and Controls Conference, October 11-13, 2017, Tysons Corner, Virginia, USA.
|
Progress 08/15/15 to 08/14/16
Outputs
Target Audience: Target Audience In this reporting period, 5 graduate students and 1 postdoc researchermajoring in the mechanical and aerospace engineering and the agriculture engineering were actively involved in the robot design enhancement and new software development for automation. Also they were involved in the testing of robots (ground and aerial) (at UCF) and the disease detection sensor (at UF). The research progress has been demonstrated to a wide range of audiences through different events: The 3nd NRI PI meeting (November 2015, Washington DC) 2016 UCF MAE graduate research day (April 2016, Orlando FL) Driscolls conference demonstration (January 2016, Tampa FL). Demonstrating the strawberry robot to 25 high school students at Hardee Senior High School 2016 ASABE international conference (July 2016, Orlando FL) Sensing, Control and Automation Technologies for Agriculture - 5th AGRICONTROL 2016 Conference (August 2016, Seattle) Also we conducted testsof the developed autonomous robot in three different commercial farms (Pappy's Patch, G&D commercial farm, and O'Brien Family Farms). Impacts or Expected Impacts The impacts or expected impacts to the communities are similar to what have been reported in previous years. This year, no senior design team was involved. However, many undergraduate students have toured the labs and some of them have showntheir interests in participating in this project. Two of the fourgraduate students at UCF working on this project are from previous undergraduate senior design teams. Conference proceedings and journal papers in the areas of robotics and precision farming have been submitted, some of which were published. The autonomous cooperative robotic system has the potential of dramatically reducing labor cost in disease detection operations. Currently the disease detection process is manual, time consuming, and can be late in detecting diseases. The disease and stress detection technique that can identify the stress at an early stage is of great interest to many growers. Efforts In the reporting period, two graduate students are Hispanic. Also two female graduate students at UCF have helped the robot and disease detector designs. Additionally Jinzhu Lu, a visiting Ph.D. student from Zhejiang University, China was hosted at UF. She worked on collecting and analyzing spectral data for major disease of tomato and strawberry as a part of this project. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?Professional training and development activities were not planned in this reporting period. How have the results been disseminated to communities of interest?Research results have been disseminated through different approaches. First, we presented our findings in many events such as 3rd NRI PI meeting, Driscolls January Demo event, 2016 UCF MAE graduate research day, and 2016 ASABE international conference, and Agricontrol 2016. Second, we submitted and published 1 journal paper, and presented 5 conference papers. Also we have submitted 2 journal papers and have one poster. Third, research results have been showcased to general public through our experiment/demo in three commercial farms (Pappy's Patch, G&D commercial farm, and O'Brien Family Farms). What do you plan to do during the next reporting period to accomplish the goals?Significant progresses have been made in the previous three years. The following tasks are planned for Year 4 (August 15, 2016 - August 14, 2017). Ground robot: Finish the development of cross-the-bed control algorithm and test it in at least one commercial farm. Test the fully integrated field navigation functionality of the robot. New path/trajectory planning algorithm will be designed and tested based on the identified robot dynamic model. Leaf sampling control system design will be finalized. Strawberry aerial robot: Real-time geo-location functionality via flight test. Disease Detection Sensor: Integrate the disease detection sensor with the ground robot Test the disease detection functionality onboard the ground robot in a commercial farm. Education, Outreach, and Dissemination: Submit journal papers and attend conference/meetings. Supervise >= 4 graduate students Support >= 1 postdoc researcher. Attract and retain students from under-represented groups. Reach out to one more growers/producers who are potentially interested in this research product. Conduct a first round cost analysis of the autonomous ground robot in disease detection scouting.
Impacts
What was accomplished under these goals?
Year 3 Project Goals and Milestones: The planned tasks in this reporting period are listed below and the approximate percentages of the accomplished tasks are listed as well. Ground robot (Goal/Object 1): Testing of the ground robot full mobility and flexibility in 1-2 commercial strawberry orchards (100%). Guidance and control of the robot through strawberry orchards with cross-bed capabilities (50%) Dynamic modeling and parametric identification of the robot with wheel-terrain interactions (100%). Leaf sampling control system via vision (70%) Strawberry aerial robot (Goal/Object 2): Flight test of the octo-rotor to enhance its stability and robustness (100%) Fight test in a commercial farm (0%) Flight test the geo-location functionality (90%) Disease Detection Sensor (Goal/Object 3): Continue the disease detection sensor design (100%) Develop the imaging/spectral analysis algorithm and software. The detection success rate should reach 60%, and the imaging processing speed should reach 0.1Hz (10 seconds processing time) (100%) Education, Outreach, and Dissemination (Goal/Object 4): Submit journal papers and attend conference/meetings (100%) Supervise >= 4 graduate students (100%) Support >= 1 postdoc researcher (100%) Attract and retain students from under-represented groups (100%) Reach out to one more growers/producers who are potentially interested in this research product (100%) Major Activities Conducted to Accomplish Goal/Object 1 There are four major activities in the ground robot development and testing. First, the ground robot was tested in commercial farms: pappy's patch (small scale), G&D farm and O'Brien Family farm (large scale). A new installation approach is used for the GPS receiver so to reduce the noise. The following two functionalities are tested in the field: over-the-bed navigation and leaf picking. Second, the simulation part of the cross-bed functionality has been accomplished. The hardware testing is still ongoing. Third, a control affine dynamics model has been developed for the field robot, which considers the terrain-wheel interaction. Vehicle and terrain parameters were identified via an extended Kalman filter. Fourth, a new inverse kinematic algorithm is developed which is suboptimal (in terms of energy consumption) and semi-analytical which can generate joint commands in real-time. The algorithm has been tested in a farm environment. It is shown that the performance index will only scarify around 6%, however with a computational cost reduction by more than 97%. Major Activities Conducted to Accomplish Goal/Object 2: There are two main accomplishments achieved in this report period. First, a detailed octorotor model is derived considering the gimbal-mounted camera. An orthogonal based projection method is used to relate pixel location of the suspected diseased leaf to the global coordinate obtained via GPS and altimeter. Second, a Kalman filter is designed and flight tested to estimate the position of the suspected diseased leaf. Major Activities Conducted to Accomplish Goal/Object 3: We achieved the maximum data collection speed of 0.5 Hz (one package of data from two spectrometer sensors NIR and VIS every two seconds). This includes automatically recognizing which sensor is connected to which universal data bus; calling of each addressed sensor; requesting data of defined wavelengths; requesting data of detected light intensity (counting number of reflected photons); transferring data packages from each sensor via Wi-Fi; data demodulation and converting from static text string to analyzable variables; and recording data into separate MS Excel *.xls and MATLAB *.fig files in specific defined folders. Therefore, the other 8 seconds remain for data processing purpose. Data collected All the design documents, testing data, and outreach and dissemination activity pictures are collected and saved. Summary statistics and discussion of results As reported previously, the main specifications of the ground robot are the same. The robot dimensions are: 91.7, 146.8, 82.6 cm. The robot can transform, expanding 50 cm and 25 cm in its Y and Z axis respectively in increments of 5 cm. The overall vehicle weight was measured to be 148 kg. The robot can perform in-site rotations, linear translations, and driving in an arc. The fastest translational speed was measured as 35 cm/sec in laboratory tests, while the speed limit is set to 10 cm/sec for field operations. The octo-rotor specifications are also the same as shown in the previous report. The weight is 8.9kg. Its full wind span is 1.5 meters. The maximum payload is 6.8kg. Estimated flight time is 23 minutes. The multi-diseased leaf detection accuracy is better than 0.5 meters, which is enough to distinguish between different rows. The disease detection unit includes five parts, which are two spectrometer sensors, signal conditioning system, wireless interface, developed data collection software, and data analysis and classification software. The spectrometer sensor part consists of two enhanced high performance linear CMOS image sensors (ELIS-1024) that cover a full visible light spectrum range of between 337 and 822 nm and the near infrared band between 634 and 1123 nm. Each sensor has a resolution of 1024 pixels with a very low noise for detecting the reflected light from an object which is exposed to the sensor. The signal conditioning system includes a mini computer and Linux local web server. The data of wavelength and the light intensity are extracted from each spectrometer sensor and produced as a text string in the server-side scripting language (PHP). The developed data collection software can remotely trigger the sensor, immediately collect data from both sensors and record them in separated Microsoft Excel files for future analysis. The data analysis software, which is currently under construction, is flexible in terms of selecting the method of data analysis for different disease and classifying the analyzed data. Key Outcomes or Other Accomplishments Realized. A change in knowledge: The key outcome will be the fully autonomous robot that can efficiently scout a commercial strawberry orchard considering scalability/adaptability, lighting control, low cost, and ability to transverse over sandy/grass terrains. Also the new detection sensor will be able to detect water stress in its early stage, which will be onboard of the ground robot and aerial platform. In this reporting period, the knowledge we gained has been presented in journal papers, conferences, and meetings. A change in action: In this year, undergraduate senior teams were not involved due to the fact that our research focus has been shifted from mainly mechanical designs to software development for autonomous operations. Therefore, graduate students and postdoc researchers were involved in design and testing. A change in condition: With further testing and enhancement, the cooperative UAV and ground robot equipped with a novel disease detection sensor will change the current, labor intensive practice in the field.
Publications
- Type:
Journal Articles
Status:
Published
Year Published:
2016
Citation:
S. G. Defterli, Y, Shi, Y. Xu, and R. Ehsani, Review of Robotic Technology for Strawberry Production, American Society of Agricultural and Biological Engineers Applied Engineering in Agriculture, Vol. 32, No. 3, 2016, pp. 301-318.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2016
Citation:
C. Garcia, Y. Xu, Target Geolocation for Disease Detection Applications via an Octorotor, the 5th IFAC Conference on Sensing, Control and Automation for Agriculture, August 14-17, 2016, Seattle, Washington, USA.
- Type:
Conference Papers and Presentations
Status:
Accepted
Year Published:
2016
Citation:
Pablo, M., Garcia, C. A., Freese, D., Defteli, S., and Xu, Y., Software and Hardware Architectures in Cooperative Aerial and Ground Robots for Agricultural Disease Detection, Accepted to the 7th International Workshop on Collaborative Robots and Human Robot Interaction, Orlando, FL, 2016.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2016
Citation:
Menendez-Aponte, Pablo, and Xu, Yunjun, Dynamic Modeling of a Skid Steer Robot used for Scouting in Strawberry Orchards 2016 ASABE Annual International Meeting, Orlando FL, July 17-20, 2016
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2016
Citation:
Garcia, Christian, Xu, Yunjun, "Onboard Strawberry Disease Detection using an UAV in Hover," 2016 ASABE Annual International Meeting, Orlando FL, July 17-20, 2016.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2016
Citation:
Defterli, Sinem G; Menendez-Aponte, Pablo; Xu, Yunjun, A Leaf Picking Algorithm for a Strawberry Disease Detection Robot, 2016 ASABE Annual International Meeting, Orlando FL, July 17-20, 2016.
- Type:
Journal Articles
Status:
Under Review
Year Published:
2016
Citation:
S. G. Defterli, Y. Xu, Analytical Solution of a Five-Degree-of-Freedom Inverse Kinematics Problem for the Handling Mechanism of an Agricultural Robot", submitted to Robotica.
- Type:
Journal Articles
Status:
Under Review
Year Published:
2016
Citation:
Lu, J., Y. She, R. Ehsani, A. Decastro, and Y. Xu. Field detection of strawberry Colletotrichum disease using hyperspectral technology. Computers and Electronics in Agriculture, Under Review.
- Type:
Theses/Dissertations
Status:
Published
Year Published:
2016
Citation:
Sinem G. Defteli Kinematic Modeling and Analytical Inverse Kinematic Solution for the Handling Mechanism of an Agricultural Robot August 2016
- Type:
Theses/Dissertations
Status:
Published
Year Published:
2016
Citation:
Chris Garcia Geolocation of Diseased Leaves in Strawberry Orchards for a Custom-Designed Octorotor August 2016
- Type:
Theses/Dissertations
Status:
Published
Year Published:
2016
Citation:
Ying She- Application of High Resolution Aerial Imaging For Inventory Management In Nurseries And Citrus Groves - August 2016
- Type:
Other
Status:
Published
Year Published:
2016
Citation:
(Poster) Freese, Douglas; Xu, Yunjun, "Navigation of Agricultural Fields using a Vision and Range Finder based Controller," 2016 ASABE Annual International Meeting, Orlando FL, July 17-20, 2016. Poster
|
Progress 08/15/14 to 08/14/15
Outputs
Target Audience:Target Audience: In this reporting period, 12 undergraduatestudents, 6 graduatestudents, and 1 postdoc researchermajoring in mechanical and aerospace engineering and agriculture engineering were actively involved in design enhancement. First, the research project has reached a wide range of audiences through different events such as (i) the 2nd NRI PI meeting (Nov. 2014, Washington DC); (ii) 2015 UCF MAE graduate research day (Feb. 2015, Orlando FL); (iii) 2014-2015 UCF MAE undergraduate senior design showcases (April 2015, Orlando FL); (iv) workshop for UAVs Applications in Agriculture, Lake Alfred, FL, March 2015; (v) annual "Posters and Pastries"Research Gallery, Lake Alfred, FL, April 2, 2015; and (vi) ASABE Annual International Meeting, New Orleans, July 2015. Second,we have demonstrated our research progress to communities through different channels. (i) We have shown how the ground robot is working in a commercial farm for general public; (ii) we have attracted many audiences in the workshop; (iii) students in the senior design team at UCF; (iv) we showed our progress in the 2nd NRI PI meeting in DC November 2014; (v) we showed our results in the ASABE meeting. Third, several news about our research project have been released to general public this year: (i) http://today.ucf.edu/ucf-researchers-working-disease-detecting-robot/; (ii) http://www.growingproduce.com/vegetables/researchers-seeking-agricultural-applications-for-unmanned-aircraft/; and (iii) http://m.theledger.com/Section/422/Article/150329749. Efforts: In the second year of the project, twelve undergraduate students were involved and two of them have been recruited into our UCF mechanical engineering master program (both of them are Hispanic). One female graduate student (at UCF), two female graduate students (at UF), and one female PostDoc researcher (at UF) helped the robot design and disease detector design, respectively. Additionally, we have one graduate student (at UCF) in his second year is also from the under-represented group (Hispanic). Same as the first year, the agriculture robots design content has been seamlessly merged into the UCF undergraduate senior design course. This year, the senior design students have written more than 4 reports to cover their preliminary design, critical/parametric design, and final design. Also they have been involved in many field tests in a commercial strawberry farm. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?
Nothing Reported
How have the results been disseminated to communities of interest?Research results have been disseminated through different approaches. (i) First, we presented our findings in different events (international, national, and local). (ii) Second, we submitted and published journal papers and conference papers. (iii) Third, research results have been showcased to general public through our experiment/demo in a commercial farm. (iv) Fourth, several news are released talking about this project. What do you plan to do during the next reporting period to accomplish the goals?The following tasks are planned for Year 3 (August 15, 2015 - August 14, 2016). Ground robot: Testing of the ground robot (full mobility and flexibility) in 1-2 commercial strawberry orchards. Guidance and control of the robot through strawberry orchards with cross-bed capabilities. Dynamic modeling and parametric identification of the robot with wheel-terrain interactions. Leaf sampling control system design via vision information Strawberry aerial robot: Fly tests of the octo-rotor to enhance its stability and robustness. Flight test its geo-location functionality. Disease Detection Sensor: Continue the disease detection sensor design Develop the imaging/spectral analysis algorithm and software. Education, Outreach, and Dissemination: Submit journal papers and attend conference/meetings. Mentor senior design teams at UCF. Supervise >= 4 graduate students Support >= 1 postdoc researcher. Attract and retain students from under-represented groups. Reach out to one more growers/producers who are potentially interested in this research product.
Impacts
What was accomplished under these goals?
As planned, the major tasks in this reporting period is listed below and the approximate percentages of the tasks that wereaccomplished are listed. Ground robot (Goal/Object 1): Testing of the ground robot in 1-2 commercial strawberry orchards. (50%) Guidance and control of the robot through strawberry orchards while avoiding obstacles. (30%) Dynamic modeling and parametric identification of the robot. (40%) Leaf sampling control system design. (60%) Strawberry aerial robot (Goal/Object 2): Fly tests of the octo-rotor to enhance its stability and robustness. (70%) Disease Detection Sensor (Goal/Object 3): Continue the disease detection sensor design (50%) Develop the imaging/spectral analysis algorithm and software. The detection success rate should reach 75%, and the imaging processing speed should reach 0.1Hz (10 seconds processing time) (70%) In addition, all the educational goals planned have been achieved. 1 Major Activities Conducted to Accomplish Goal/Object 1 There are four research components studied in this reporting period. First,in order to fit the robot within a typical commercial orchard, a "transformer" type structure has been designed such that the height and width of the robot can be adjusted for different strawberry fields.For the robot to be able to travel in different terrain conditions, the original 6-wheel design has been changed to a4-wheel design and a more powerful powertrain has been incorporated. Second, algorithm and software have been designed and implemented so that the robot can easily travel over beds without collision with the beds.Information among two Arduino board, laptops, and sensing and actuation devices have beenre-designed. Part of the vision algorithms are currently under-development. The trajectory planning algorithm has been published in a journal. Third, the task associated with the robot modeling and identification is still on-going. Fourth, the forward kinematics and a new inverse kinematic of the leaf sampling control algorithm are developed and tested in simulation. 2 Major Activities Conducted to Accomplish Goal/Object 2: First, the power bus has been redesigned so it can provide more reliable and robust power supplies to all the subsystems. Second, the landing gear has been redesigned. A conventional landing gear requires complex guidance and control systems so that the octo-rotor can land safely in different weather and terrain conditions.Here a special landing gear with a large coverage area has been used. Third, a protection subsystem has been developed. If the octo-rotor is too close to the ground, the box is closed. Fourth, the geo-location algorithm has been developed to estimate the location of suspected diseased leafs. 3Major Activities Conducted to Accomplish Goal/Object 3: During the second year, the activities focused on evaluating the feasibility to use hyperspectral and thermal in-field data for an accurate detection of Colletotrichum disease in strawberry plants as well as the drought stress detection in some specialty Florida crops. 3.1 Hyperspectral analysis in strawberry disease detection Hyperspectral data from healthy as control, symptomatic-infested and asymptomatic-infested leaves were collected under field conditions using a developed mobile platformon a commercial strawberry farm located in Plant city, FL.A handheld spectroradiometer (SVC HR-1024, Spectra Vista Cooperation, NY)is on the platformwith a 4º field-of-view optical lens in the spectral range of 350 to 2,500 nm. Indoor spectral data under a controlled environment were carried out based on sample leaves collected from the same plants for comparison with field ones. Five typical varieties were used in this study: sensation, festival, pilgrim, radiance and Sanibel. This study demonstrates the possibility of using hyper-spectral technology based on spectral vegetation indices (SVIs) for strawberry Colletotrichumdisease. More analysis should be done to find the optimum conditions and algorithm to improve those classifications. 3.2 Thermography and hyperspectral sensing data for Citrus and Strawberry Stress Detection Experiment were conducted on a commercial nursery in Virginia Beach, VA, and experiment fields in the Citrus Research and Education Center, Lake Alfred, Fl. A commercial thermal camera and a customized low-cost multi-pixel thermometer (Infrared Cameras Inc., TX) were compared for their performance on the drought stress detection on some specialty crops: citrus, strawberry, rose and barberry. Due to the dependence of canopy temperatures on environment factors, air temperature, air relative humidity, photosynthetically active radiation (PAR) and wind speed data were also collected. The preliminary study results demonstrated the feasibility of drought stress detection using open-field thermal sensing and indoor hyperspectral sensing on specialty crops including strawberry and citrus. The technologies can be integrated with the field robot sensing platforms to conduct more systematic tests. 4Data collected All the design documents, testing data, and outreach and dissemination activity pictures are collected and saved. 5Summary statistics and discussion of results 5.1 The robot dimensions are: 91.7, 146.8, 82.6 cm. The robot can transform, expanding 50 cm and 25 cm in its Y and Z axes respectively in increments of 5 cm. The overall vehicle weight was measured to be 148 kg. A 35 Ah pair of 12 V batteries powers the drivetrain with an estimated travel distance of 8 miles.The drivetrain navigation accuracy was measured as approximately 3 cm for the robot travel distance in its X axis. In field testing driving straight down a strawberry row, the lateral offset in the Y axis was measured as -4.1 cm considering the uneven terrain and travel distance of 10 meters; this is significant.The fastest translational speed was measured as 35 cm/sec in laboratory tests, while the speed limit is set to 10 cm/sec for field operations. 5.2 The UAV weight is 8.9kg. Its full wind span is 1.5 meters. The landing gear area is 2.5 m2. The height of the vehicle is 0.73 m and the maximum payload is 6.8kg. The estimated flight time is 23 minutes. The battery voltage is 22.2 volts. Its altitude hold vertical acceleration is 250 cm/s2, and the waypoint accuracy is 2 meters. 5.3 Disease detection sensor research: (i) relationship was established between the canopy temperature predicted by the measured environment factors and the measured canopy temperature by the thermal camera;(ii) for the strawberry plants with six different irrigation levels, a correlation with R2 of 0.68 was achieved between the measured canopy temperature and the average stomatal conductance; (iii) good correlations (0.54 - 0.90) were achieved between the customized low-cost multi-pixel thermometer and the commercial thermal camera; and (iv) good correlations (0.82 - 0.97) were achieved between vegetation indices NDVI, NDWI, SRI and PSRI calculated for rose plants with different water stress levels and their irrigation amounts. 6Key Outcomes or Other Accomplishments Realized. A change in knowledge: The knowledge we gained has been presented in journal papers, conferences, meetings, and news. After two years of development, a good basis for developing an automatic disease detection system is laid out. A change in action: In this year, the project further enhances students' learning effectiveness through hand-on design, manufacturing, and experiment activities, which also improved their understanding about the materials in courses. In addition, different from Year 1, graduate students are actively participating in supervising undergraduate design teams. A change in condition: The cooperative UAV and ground robot equipped with a novel disease detection sensor is an innovative approach, as compared with the existing manual process.
Publications
- Type:
Journal Articles
Status:
Published
Year Published:
2015
Citation:
" N. Li, C. Remeikas, Y. Xu, S. Jayasuriya, and R. Ehsani, Task Assignment and Trajectory Planning Algorithm for a Class of Cooperative Agricultural Robots, the ASME Journal of Dynamic Systems, Measurement, and Control, Vol. 137, No. 5, May 01, 2015, pp. 051004.
- Type:
Journal Articles
Status:
Under Review
Year Published:
2016
Citation:
S. G. Defterli, Y, Shi, Y. Xu, R. Ehsani, Review of Robotic Technology for Strawberry Production, ASABE ASE, July 2015, revision submitted
- Type:
Other
Status:
Other
Year Published:
2014
Citation:
(Poster presented in the 2nd NRI PI meeting) USDA/NRI: Automated stress and disease detection in vegetable and tree crops using a cooperative ground and aerial network November 2014.
- Type:
Journal Articles
Status:
Other
Year Published:
2016
Citation:
D. Freese, S. G. Defterli, P. Menendez-Aponte, Y. Xu, and R. Ehsani, Design of an agricultural robot for disease detection in strawberry orchards, to be submitted.
- Type:
Conference Papers and Presentations
Status:
Under Review
Year Published:
2016
Citation:
C. Garcia, and Y. Xu, Target Geolocation for Agricultural Disease Detection Applications via an Octorotor, submitted to the 2016 AIAA SciTech conference.
- Type:
Other
Status:
Other
Year Published:
2015
Citation:
J. Lu, Y. Shi, J. Abdu-Ridha, Y. Xu, R. Ehsani, Hyperspectral Sensing on Strawberry Colletotrichum Disease. ASABE Annual International Meeting, New Orleans, July 25-29, 2015. (Oral presentation)
- Type:
Other
Status:
Other
Year Published:
2015
Citation:
Y. Shi, Y. She, J. Lu, R. Ehsani, J. Owen. Ground-based water stress detection on nursery ornamentals using thermography and hyperspectral technologies. ASABE Annual International Meeting, New Orleans, July 25-29, 2015. (Oral presentation)
- Type:
Other
Status:
Other
Year Published:
2015
Citation:
Shi, Y., Y. She, and R. Ehsani. Comparison of a commercial thermal camera and a customized thermometer on plant water stress detection. Poster presented at the 4th annual Posters and Pastries Research Gallery, Lake Alfred, FL). April, 2015. Received best poster Awarded.
- Type:
Other
Status:
Other
Year Published:
2015
Citation:
Yeyin Shi. Water stress detection using thermal sensing technology . Presented in Workshop for UAVs Applications in Agriculture, Lake Alfred, FL, March 2015.
- Type:
Journal Articles
Status:
Other
Year Published:
2016
Citation:
J. Lu, Yeyin She, R. Ehsani, A. Decastro, Y. Xu. Field detection of strawberry Colletotrichum disease using hyperspectral technology. Under internal Review.
|
Progress 08/15/13 to 08/14/14
Outputs
Target Audience: Target Audience In the first year of the project, about 12 undergraduate and 3 graduate students majoring in mechanical and aerospace engineering and agriculture engineering were actively involved in the design, building, and testing of the robots (ground and aerial) and the disease detection sensor. First, the research progress in developing the networked autonomous systems to enhance the efficiency in stress and disease detection operations has reached a wide range of audiences through conferences, expo, and showcases in different communities. Specifically, the research results have been presented or shown in •The 1st NRI PI meeting (Oct. 2013, Washington DC) •2014 RHEA meeting (May 2014, Madrid, Spain) •2014 AUVSI’s unmanned systems (May 2014, Kennedy Space Center, FL) •2014 UCF MAE graduate research day (Feb. 2014, Orlando FL) •2013-2014 UCF MAE undergraduate senior design showcases (April 2014, Orlando FL) Second, we have reached out to local school students and teachers in their field trip and summer camp. During our field test in the UCF Arboretum garden, a group of kids from a local elementary school was attracted (June 2014). Also our developed agriculture robots (both aerial and ground) were introduced to more than forty 8th-10th grade students who were attending the UCF Campus Connect program (July 2014). Third, the project progress and overall goals have been released to general public through media coverage. In the first year, the CECS news, UCF Today, Orlando Business Journal, and Central Florida Future have reported our project news. Impacts or Expected Impacts The impacts or expected impacts brought to the audiences mentioned above are summarized here: •It is the first time at UCF, an emerging agriculture related project is introduced into the Mechanical / Aerospace Engineering curriculum, which has and will continue to inspire many students. They are aware of new opportunities that can influence their careers. •The innovative ground robot and octo-rotor design and their hardware demonstration have attracted much attention from a wide range of audiences in academia, industry, growers/producer, and general public. •The research results have and will be published or presented to professionals in the communities of robotics, control engineering, and precision farming. •Media coverage has helped us disseminate new technologies under development to agriculture industry as well as general public. •The disease and stress detection technique under development that can identify the stress at the early stage is of great interest to growers. Efforts In the first year of the project, several students involved in the project are coming from under-represented group. For example 1 African American student and 1 Hispanic undergraduate student are contributing to the design, development, and test of the ground and aerial robots. Additionally, one female graduate student (at UCF) and one female PostDoc researcher (at UF) are helping the robot design and disease detector design, respectively. The agriculture robot design content has been seamlessly merged into the UCF undergraduate senior design course. The control and optimization topics have been introduced in the graduate level optimal control course. The research project has generated one master thesis (Charles Remeikas, Dec. 2013). Also senior design students involved in this project have written more than 6 reports to cover their preliminary design, critical/parametric design, and final design. Those reports are available if requested by USDA. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided? Professional training and development activities were not planned in the first year. How have the results been disseminated to communities of interest? Research results have been disseminated through different approaches. First, we presented our findings in five different events (international, national, and local): 1st NRI PI meeting, 2014-RHEA meeting, 2014 AUVSI unmanned systems, 2014 UCF MAE graduate research day, and 2013-2014 UCF MAE senior design final show. Second, we submitted and published two journal papers, and 1 conference paper. Among them, one journal (under review) and 1 conference are in the community (Dynamics and Controls in the Mechanical Engineering) where traditionally agricultural robot topics are related but largely missing. Third, research results have been showcased to elementary and middle school students. Fourth, several news are released talking about this project. Fifth, lots of interest has been generated in the research community about using or designing innovative autonomous systems to enable efficient and precision farming. What do you plan to do during the next reporting period to accomplish the goals? To meet the overall goal and realize the anticipated impacts, three main tasks are planned in the project: sensing and monitoring system, real-time optimal decision making framework, and experiment validation. Promising results have been obtained in the first year. The following are the tasks and objectives to be accomplished during the next reporting period, which are categorized into strawberry ground robot, aerial UAV, disease detection sensor, communication, and education and dissemination. Ground robot: •Testing of the ground robot mobility and flexibility in 1-2 commercial strawberry orchards. •Guidance and control of the robot through strawberry orchards while avoiding obstacles. •Dynamic modeling and parametric identification of the robot. •Leaf sampling control system design. Strawberry aerial robot: •Fly tests of the octo-rotor to enhance its stability and robustness. Disease Detection Sensor: •Continue the disease detection sensor design •Develop the imaging/spectral analysis algorithm and software. The detection success rate should reach 60%, and the imaging processing speed should reach 0.1Hz (10 seconds processing time) Education, Outreach, and Dissemination: •Submit journal papers and attend conference/meetings. •Mentor senior design teams at UCF. •Supervise >= 3 graduate students •Support >= 1 postdoc researcher. •Attract students from under-represented groups. •Reach out to growers/producers who are potentially interested in this research product.
Impacts
What was accomplished under these goals?
Potential Impact: Early and efficient disease infection management is a critical task in fruit (e.g. strawberry and citrus) productions as it affects fruit quality, yield, and ultimately market value. Costly and labor intensive manual scouting is currently the most frequently used method of disease and stress detection in crops. If the networked aerial and ground robot system equipped with the new disease detection sensor is successfully developed, commercialized, and adopted by producers/growers, it could potentially reduce their scouting costs by 50% and improve the accuracy of stress detection by 30%, which could significantly increase their profits. Additionally, if the project is successful, further interests in precision farming and agriculture automation using robotic technologies are expected. Project Objectives: The proposed research tasks leverage strong experiences in cooperative robots' decision making at UCF and citrus disease detection research conducted at UF to reach the ultimate goal of reducing scouting costs and crop loss due to several different stress factors. Specific Goals: A robust aerial crop health monitoring system is in the process of generating geo-referenced crop health maps and high level fruit-specific formation configurations (year 1 to year 3). These maps are then to be supplied to the cooperative agricultural ground platform (year 1 to year 4) equipped with novel sensing tools to navigate to the aerially-diagnosed plant/tree locations and perform detailed sensing tasks. The integrated system will then be evaluated extensively in Citrus Research and Education Center (CREC) research farms under realistic field conditions followed by overall modification to the network system based on previous findings (year 2 to year 4). The enhanced system will then be tested and validated in commercial farms/orchards (year 4 to year 5). Year 1 Project Goals and Milestones: To meet the overall project goals, the detailed milestones have been set for each of the five years. In the first year, the following planned project tasks and milestones have been accomplished. •Develop an innovative ground robot that will be used to monitor and analyze a strawberry field for diseases in close-proximity by taking spectral imaging and collecting leaf samples throughout the field. •Develop an octo-rotor that will be used to monitor and analyze a strawberry field rapidly and relay the suspected area coordinates to the ground robot. •Develop the first version of a water stress detection sensor, which could be applied on both the aerial- and ground-based platforms, and conducted preliminary field tests. •Disseminate research results to a broader range of audience. Major Activitiesfor Object 1 A team of six undergraduate students and two graduate students, supervised by the PI, developed a new ground robot according to the specifications of a typical strawberry farm. The design process includes the engineering design specification, preliminary design, critical design, and final design. After the design is accomplished, all the parts are manufactured/purchased, integrated, and finally tested in the UCF Arboretum garden. Major Activitiesfor Object 2: Six undergraduate students designed and built a new octo-rotor, which can carry the new disease detection sensor (under development) and rapidly survey across vast fields. The students have gone through the same training and mentoring process as those in the ground robot team. The designed octo-rotor can carry 5-10 lb and fly at least for 20 minutes. Major Activities for Object 3: Water stress is one of the major stresses in the production of strawberries and citrus. The core component of the sensor system is a miniature low-cost yet accurate infrared (IR) sensor. It is so small that could be integrated in a compact circuit board and carried with ground or aerial platforms. The hardware and software were developed to realize functions including communicating with the sensor, extracting temperature readings, and storing data on a micro SD card. The onboard program directly converted the raw package data to Celsius temperature readings in a frequency of 4 Hz. A GPS was included in the system so that each temperature reading was stored with location information for the later geo-referencing. Major Activities for Object 4: In the first year, the project director (Dr. Xu, UCF) and co-project director (Dr. Ehsani, UF) supervised 12 undergraduate students (at UCF), 3 graduate students (2 at UCF and 1 at UF), and 1 postdoc researcher (at UF). One graduate student under Dr. Xu’s supervision who is contributing partially to this project graduated with a master degree. We presented our project in conference meetings and UCF MAE events. The robots are showcased at the 2014 AUVSI’s unmanned systems expo on May 2014. Our project was covered by online news outlets. Data collected: All the design documents, testing data, and outreach and dissemination activity pictures are collected. For example, the collected data in the sensor was temperature readings in Celsius unit and GPS data including locations and time. Summary statistics and discussion of results: The specifications of the ground robot are: length (91.7 cm), width (146.8cm), height (82.6cm), wheel base (33cm),and degrees of freedom (DOF: total 8 excluding the cutting device and the curtain control device). The specifications of the octo-rotor are: weight (without payload) (8.8kg), full wingspan (1.5m), height (0.5m), wheel base (33cm), and estimated flight time when carry a 6.8kg payload is about 22 minutes. The sensor calibration in the lab revealed a good correlation between the sensor-measured temperatures with the actual temperatures (R2 > 0.99). The preliminary field tests showed the sensor’s ability to differentiate the vegetation area from the soil which had a clear temperature difference. Key Outcomes or Other Accomplishments Realized: A change in knowledge: The uniqueness of the vehicle strongly tightened with the strawberry application includes: scalability/adaptability, lighting control, low cost, and ability to transverse over sandy/grass terrains. The knowledge we gained has been presented in journal papers, international conference, meetings, news, and UCF MAE events. Also this first-year sensor study established a good basis for developing an automatic water stress detection system for aerial and ground agricultural robots. A change in action: This project enhances students’ learning effectiveness through extensive hand-on design, manufacturing, and experiment activities, which also improved their understanding about the materials in courses such as Flight Mechanics, Analysis and Design of Aerospace Structure, Fundamental of Aerodynamics, Mechatronics, and CAD/CAM. The current study aims to find the best way to apply low-cost thermal sensing technology to detect water stress. A change in condition: The ultimate goal of the current study aims to substitute heavy manual scouting for crop stresses with an automatic, timely, and accurate sensing system. Its accomplishment will be useful not only in replacing labor-cost manual scout, but also in scheduling an irrigation plan to reach an optimal water supply with a maximized yield and a minimized production cost for farmers, and further benefit the water conservation and world hunger in an international scope. If the project is adopted by strawberry producers, it will reduce labor cost and increase profits. The designed octo-rotor can be used in not only our target strawberry/citrus farms but also other agriculture applications. Students gained significant training and preparation of their future career. Robot demonstrations attracted elementary and middle school students and hopefully inspired them to pursue their education and career in STEM fields.
Publications
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2014
Citation:
Y. Xu, R. Ehsani, J. Kaplan, I. Ahmed, W. Kuzma, J. Orlandi, K. Nehila, K. Waller, and S. Defterli, An octo-rotor ground network for autonomous strawberry disease detection Year 1 Status Update, the 2nd International Conference on Robotics and Associated High-Technologies and Equipment for Agriculture and Forestry, May 21-23, 2014, Madrid, Spain, pp. 457-466.
- Type:
Journal Articles
Status:
Under Review
Year Published:
2013
Citation:
N. Li, C. Remeikas, Y. Xu, S. Jayasuriya, and R. Ehsani, Task assignment and trajectory planning algorithm for a class of cooperative agricultural robots, Submitted to the ASME Journal of Dynamic Systems, Measurement, and Control.
- Type:
Journal Articles
Status:
Published
Year Published:
2013
Citation:
K., Johnson, S. Sankaran, and R. Ehsani, Identification of water stress in citrus leaves using sensing technologies, Agronomy 3(4), November 2013, pp. 747-756.
- Type:
Other
Status:
Other
Year Published:
2013
Citation:
Poster in the 1st NRI PI Meeting: "Automated stress and disease detection in vegetable and tree crops using a cooperative ground and aerial network October 1-2, 2013
- Type:
Theses/Dissertations
Status:
Published
Year Published:
2013
Citation:
C. Remeikas, Bio-inspired Cooperative Optimal Trajectory Planning for Autonomous Vehicles, Master Thesis, Dec. 2013
|
|