Progress 09/01/23 to 08/31/24
Outputs
Target Audience:During this reporting period, Birds Eye Robotics primarily targeted commercial poultry integrators, strategic partners in animal health, and poultry farm operators, focusing on those interested in improving poultry welfare, operational efficiency, and labor reduction through automation. The core audience includes farm management teams and poultry production stakeholders who face labor shortages and are actively seeking technology-driven solutions to improve both animal welfare and barn productivity. Birds Eye Robotics' collaboration with Merck Animal Health, a strategic partner with a strong focus on data-driven welfare analytics, played a pivotal role in reaching this target audience. Merck Animal Health's expertise in animal health and welfare analytics has provided valuable insights into how autonomous technologies, like the Caretaker Robot, can contribute to the broader goals of welfare enhancement and operational optimization within the poultry sector. Sharing interim project findings with Merck has allowed Birds Eye Robotics to deepen the integration of analytics and real-time data on poultry behavior, health indicators, and environmental factors, thereby extending the impact of the Caretaker Robot to a larger industry audience with aligned goals in welfare and efficiency. Poultry integrators and farm operators are targeted for their role as decision-makers who determine the adoption of new technologies that address labor shortages, improve barn conditions, and reduce costs. These operators benefit from experiential learning opportunities as they observe and engage with the Caretaker Robot's capabilities in real-time field trials. Through Robotics-as-a-Service (RaaS) leasing and field trials, Birds Eye Robotics ensures that these stakeholders experience firsthand the value of integrating automation into barn management, emphasizing the practical, scalable benefits of the Caretaker Robot across multiple farm settings. This targeted approach enables Birds Eye Robotics to reach audiences who have a direct interest in data-driven welfare improvements and automation, ensuring the technology meets specific operational needs while addressing broader poultry industry challenges. Changes/Problems: Challenges with Visual SLAM Navigation and Transition to Lidar Inertial Odometry (LIO) Initially, the Caretaker Robot relied on a Visual Simultaneous Localization and Mapping (Visual SLAM) system for barn navigation. However, Visual SLAM proved inadequate due to the inconsistent lighting and dust conditions in poultry barns. Low lighting toward the end of flock cycles and high brightness near windows often caused the system to misinterpret the environment, leading to navigation errors and system instability. Change in Approach: Based on these limitations, Birds Eye Robotics transitioned from Visual SLAM to a Lidar Inertial Odometry (LIO) system, which is more resilient to environmental inconsistencies. LIO leverages LiDAR and inertial measurement data for highly accurate navigation without reliance on visual cues, making it far better suited to the barn environment. Impact and Resolution: This shift significantly improved localization accuracy, with testing demonstrating a consistent accuracy within 15 cm variance. The change required additional time for system recalibration and testing, but it has enhanced overall reliability, supporting continuous and autonomous barn navigation. Supply Chain Issues with LiDAR Sensors The availability of the Livox Mid360 LiDAR unit, a critical component for navigation, became a challenge as supply chain issues caused delays and price increases. The Livox Mid360 is relatively affordable at approximately $750 per unit, but comparable alternatives cost closer to $3,000, impacting the budget and system design considerations. Change in Approach: The team explored alternative suppliers and potential backup LiDAR options, while prioritizing Livox inventory to ensure project continuity. Efforts are ongoing to source additional units at a manageable cost and to identify alternative suppliers to mitigate future disruptions. Impact and Resolution: While budget allocations required adjustments, alternative sourcing options have helped keep the project on track. This change may require additional budget planning if supply issues persist, but proactive sourcing has minimized any immediate delays in testing or deployment. Underutilization of Battery Capacity and Adjustments to Charging Protocols Initial field tests revealed that the Caretaker's battery was not fully utilized in its current mission profile. The 24V 50Ah battery often retained significant charge after the completion of the robot's daily laps, highlighting an opportunity to adjust both battery size and recharging protocols to reduce costs and increase efficiency. Change in Approach: To prevent battery depletion issues, Birds Eye Robotics adopted a recharging protocol that initiates charging when the battery level reaches 75%, ensuring the robot can continue operating without risk of complete drainage. The team is also exploring smaller battery configurations and faster recharge cycles to optimize cost and performance. Impact and Resolution: The change in recharging protocol has eliminated the risk of operational downtime due to full depletion. Additionally, the potential use of a lower-capacity battery is being tested, which could reduce the bill of materials (BOM) costs for future units. These adjustments have improved both efficiency and cost-effectiveness without impacting the robot's functionality. Refinement of Mortality Detection Model Due to Lighting Challenges The Caretaker Robot's ResNet50-based computer vision model initially faced difficulties accurately distinguishing between deceased birds and bright reflections caused by LED lighting in the barn. These lighting artifacts led to false positives, which interfered with the robot's navigation and mortality detection functions. Change in Approach: Birds Eye Robotics addressed this by augmenting the ResNet50 model with additional training data focused on recognizing and filtering out lighting artifacts. This iterative refinement process required additional data collection and model testing to ensure reliable mortality detection under varying lighting conditions. Impact and Resolution: The enhanced ResNet50 model now differentiates effectively between birds and lighting reflections, significantly reducing false positives. Although this adjustment required additional training and testing time, the resulting improvement in detection accuracy has strengthened the robot's overall reliability. Operational Challenges in Field Support and Robot Operator Costs During pilot tests, Birds Eye Robotics encountered higher-than-anticipated field support and operator costs. Variability in operator hourly rates, increased customer demand for additional laps per day, and minor instabilities in the automation stack all contributed to elevated support costs. Change in Approach: To address operator cost variability, Birds Eye implemented more structured scheduling practices and increased engagement with a managed service provider. Additionally, improvements in the localization and navigation systems have contributed to better system stability, reducing the need for operator intervention. Impact and Resolution: These changes have reduced operational costs while allowing the project to meet customer demands for additional operational capacity. Better cost management practices and improved system reliability will facilitate long-term scalability. What opportunities for training and professional development has the project provided? Mentorship in Robotics and Automation: Senior robotics engineers mentored junior team members throughout the development process, specifically in areas such as LiDAR Inertial Odometry (LIO), localization techniques, and autonomous navigation. Through one-on-one sessions and hands-on guidance, junior engineers gained experience with LIO systems and were able to troubleshoot and refine localization technology, a critical aspect of autonomous navigation in unpredictable barn environments. Computer Vision Model Development: Team members working on the ResNet50 computer vision model received advanced training in machine learning and model optimization, specifically to address the challenges of mortality detection in variable barn lighting conditions. Training sessions included workshops on data annotation, model refinement, and testing in low-light environments. This training improved the team's ability to develop and fine-tune machine learning models for real-time applications, skills that are applicable in various computer vision applications across industries. Battery and Power Management Training: Given the importance of optimizing battery usage for efficient operation, team members were trained in power management strategies and wireless charging technologies, specifically with the WiBotic charging system. Hands-on training sessions covered the intricacies of battery capacity assessment, charging cycles, and protocols to prevent full depletion. This knowledge supports future applications in energy-efficient robotics design, where battery optimization is critical for operational success. Professional Development through Conferences and Industry Networking: Project leads and key team members attended robotics and agricultural technology conferences, where they had the opportunity to engage with industry experts and learn about emerging trends in agricultural automation and animal welfare technologies. These conferences provided valuable insights into best practices, allowed team members to exchange ideas with peers in the field, and facilitated potential collaborations with other agtech innovators. One example is participation at the Heartland Robotics Conference. How have the results been disseminated to communities of interest?
Nothing Reported
What do you plan to do during the next reporting period to accomplish the goals?The end period for the Phase II grant is 08/31/2025, at this time the final technical report and all reports will be completed.
Impacts
What was accomplished under these goals?
Objective 1: Reliable Localization and Navigation System Activities and Experiments Conducted: Birds Eye Robotics developed a Lidar Inertial Odometry (LIO) system to overcome the challenges of barn navigation, such as inconsistent lighting and dust. The initial testing phase confirmed the LIO system's ability to maintain localization accuracy within a 15 cm variance, surpassing previous Visual SLAM methods that struggled under barn conditions. Data Collected: Localization variance data collected during lab and initial barn tests demonstrated a consistent accuracy of under 15 cm, with only a 3 cm deviation in controlled environments. Results and Impact: The LIO system's reliability in challenging barn conditions directly contributes to the Caretaker's autonomous capabilities, ensuring effective barn coverage and minimizing reliance on human intervention for navigation. This advancement enables farmers to confidently adopt robotic solutions that operate reliably in real-world settings, reducing the risk of operational disruptions. Objective 2: Mortality Indication and Avoidance System Activities and Experiments Conducted: A ResNet50-based computer vision model was developed and refined to detect and avoid deceased or unresponsive birds. Additional training data was added to address challenges with LED lighting reflections that initially caused false positives. Data Collected: Detection accuracy and false-positive rates were recorded before and after model refinement, with significant improvements observed in the updated model's ability to differentiate between birds and lighting artifacts. Results and Impact: This accomplishment ensures that the Caretaker can safely navigate around live and deceased birds, automating a critical aspect of barn management that typically requires human oversight. For poultry operators, this feature not only improves barn safety but also promotes animal welfare by reducing the need for workers to manually monitor mortality, freeing up labor resources for other essential tasks. Objective 3: Route Planning and Battery Optimization Activities and Experiments Conducted: The project team optimized route planning and battery usage to ensure the Caretaker's efficiency across a standard 600 ft x 30 ft barn. Experiments confirmed that a 12 Ah charge supports the completion of five full laps, with a WiBotic wireless charging system recharging the battery within five hours under optimal alignment. Data Collected: Battery consumption data was collected at various points to assess efficiency across laps, and a recharge protocol was developed to prevent full depletion. Results and Impact: The optimized battery and route planning ensure that the Caretaker can operate continuously, addressing labor gaps in barn management. For poultry operators, this accomplishment means reliable daily operations without frequent downtime or manual battery management, making the system practical and efficient in real-world use. Objective 4: Testing Spinner RPM for Flock Management Activities and Experiments Conducted: Testing was conducted to determine the ideal spinner RPM for different bird age groups, with speeds of 5, 10, and 15 RPM evaluated based on flock response and battery consumption. Results showed that younger birds responded quickly at 5 RPM, while older birds required higher speeds to move effectively. Data Collected: Behavioral response data by age group and battery consumption statistics for each RPM setting were collected to inform optimal settings for energy-efficient flock management. Results and Impact: The customized RPM settings enable the Caretaker to balance power usage with effective flock redistribution, improving barn conditions while conserving battery life. This accomplishment allows poultry operators to manage flock density and movement without manual intervention, promoting healthier barn environments and reducing labor needs. Key Outcomes and Overall Project Impact Through advancements in autonomous navigation, mortality detection, route planning, and flock management, the Caretaker Robot project has made significant progress toward its goal of revolutionizing poultry barn management. These outcomes directly address critical labor shortages in the industry, enabling operators to maintain high animal welfare standards with less manual intervention. Additionally, the improved ResNet50 model and LIO navigation system underscore the project's success in adapting cutting-edge technology for robust performance in challenging barn environments. Overall, the accomplishments of this project support a sustainable and scalable approach to poultry management, providing a model for integrating robotics into agriculture. This impact is not only beneficial for poultry operators but also for the broader community, as it advances efficient, welfare-oriented farming practices essential to a resilient food supply chain.
Publications
|