Progress 07/01/24 to 02/26/25

Outputs
Target Audience:The target audiencesof this project are county and municipal governments seeking cost-effective road weather information tools for rural areas. By providing these tools, the project aims to enhance decision-making capabilities, reduce labor costs, and improve overall roadway safety, ensuring more efficient and informed responses to weather-related challenges. Changes/Problems:Our project proceeded as planned without significant challenges or problems in approach. However, following our research, we refined our product development strategy. Initially, we intended to integrate weather sensors directly into the EdgeCam. Through our findings, we determined that the most accurate and cost-effective solution was to enable the EdgeCam platform to integrate with data from existing environmental sensors already utilized by county and municipal governments. This adjustment enhances compatibility, reduces costs, and maximizes the use of current infrastructure. What opportunities for training and professional development has the project provided?This project delivered targeted training and professional development focused on the unique needs and constraints of rural county and municipal governments, current environmental sensor technologies and machine learning applications utilized by these and similar entities, and available opportunities to enhance the EdgeCam by integrating the most cost-effective and efficient sensors and machine learning models that would ensure greater functionality and value for end users. How have the results been disseminated to communities of interest?Communities of interest have been offered the opportunity to help field test an improved EdgeCam in Phase II and have been given the opportunity to weigh in on proposed updates to the current EdgeCam that would most effectively fit their needs. 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 work accomplished under these goals is as follows: Objective 1 Afterstakeholder interviews, Barn Owl conducted a comprehensive analysis to ascertain the essential sensors required to provide rural municipalities with cost-effective solutions for precise, real-time data. This research sought to harmonize affordability with functionality, enabling local agencies to make effective, informed decisions. Based on this analysis, we identified priority sensors, including road temperature, humidity, wind speed, precipitation, and hydrologic data. Appendix B provides a detailed summary of the sensor data and its applications. Barn Owl developed initial strategies to improve data access in remote areas by integrating with systems like Starlink and cellular IoT solutions. We successfully applied for and received AT&T FirstNet access, ensuring prioritized connectivity for emergency response teams and municipal agencies. Additionally, we are securing Verizon Frontline access to bolster network reliability further and expand coverage in underserved locations. Objective 2 After further review and customer feedback, it became evident that integrating sensors directly into EdgeCam for a singular data backhaul was not the preferred approach. Instead, customers emphasized the importance of a flexible, cloud-based integration model, allowing Barn Owl to implement a wider variety of sensors to deliver a more comprehensive data picture. This shift in perspective highlighted the need for a system capable of aggregating and analyzing multiple data streams at the cloud level, providing a broader and more adaptable solution for municipal and enterprise users. To address this, we prioritized the development of a flexible hardware architecture that supports seamless cloud-level integration. This approach ensures that various sensors, regardless of manufacturer or protocol, can transmit data efficiently to a centralized cloud system where advanced analytics and machine learning models can process and interpret it. This flexible architecture supports current operational needs and allows for future expansion and enhancements as technology and data requirements evolve. Objective 3 Research into existing machine learning models has provided valuable insights into the potential for automated decision-making in road and environmental monitoring. Key findings indicate that leveraging AI frameworks such as TensorFlow and OpenCV can significantly improve object recognition and pattern detection, particularly in dynamic environments. Additionally, edge AI models offer a promising approach to optimizing performance in remote areas with limited bandwidth, ensuring efficient processing without reliance on high-bandwidth transmissions and cloud computing. These insights highlight the potential for integrating advanced machine learning techniques to enhance predictive analytics, automate hazard detection, and improve overall decision-making for municipalities. Neural networks were examined for their potential in predictive analytics, particularly in forecasting hazardous conditions and identifying maintenance needs before issues arise. Building upon our findings in machine learning applications, recent research into road condition AI models has revealed significant opportunities for improving real-time monitoring and predictive analytics that can complement existing object recognition systems by analyzing visual indicators of pavement deterioration, pothole formation, ice accumulation, and surface water levels, offering municipalities a more comprehensive approach to road safety and maintenance. Various AI-based road condition assessment models have been explored, focusing specifically on the impact of weather-related road surface conditions. These modelsenablemore precise and proactive winter road treatment strategies andare able tohelp municipalities allocate resources more efficiently during snowstorms. Deep learning frameworks like Mask R-CNN have demonstrated promise in segmenting and categorizing road surface conditions affected by snow and ice. These systems analyze road images in real-time, allowing municipalities to deploy de-icing treatments and optimize plowing operations quickly.By leveraging machine learning algorithms trained on historical and real-time sensor data, municipalities can anticipate hazardous conditions, improve plowing and de-icing operations, and optimize the use of salt and other treatment materials. These advancements can significantly enhance decision-making processes by prioritizing winter road maintenance efforts, reducing accident risks, and improving overall transportation efficiency during severe weather events. A key learning from this research is the opportunity to implement AI-driven event prioritization, which has shown potential to enable municipalities to receive tiered alerts based on urgency and safety impact. While existing AI models provide a foundational capability for prioritization, further development and refinement are needed to ensure real-world accuracy and reliability. Current models can classify and rank hazards based on predefined parameters, but additional testing and training are required to enhance adaptability to evolving environmental and infrastructure conditions. Objective 4 EdgeCam units were strategically deployed in two distinct locations: St. Louis County, Minnesota, and Palo Duro Water District, Texas. These locations were chosen to assess the system's performance under diverse real-world conditions. The evaluation's primary focus was on system resilience, power efficiency, and data transmission reliability across a wide range of weather conditions. To ensure that all seasons and potential weather events were considered, the testing period was scheduled to last at least one year, with a projected conclusion in the fall of 2025. This extended timeframe would allow for collecting comprehensive data on the system's performance under varying temperatures, precipitation levels, and sunlight exposure. Preliminary feedback from the testing entities in both St. Louis County and Palo Duro Water District has been generally positive. Stakeholders have expressed satisfaction with the overall functionality and effectiveness of the EdgeCam units. However, some minor areas for improvement have been identified. Specifically, recommendations have been made to enhance the solar panel mounting system and to refine certain user features and firmware functions for a more seamless and intuitive experience. The insights gained from this real-world testing phase will be invaluable in informing future iterations of the EdgeCam system. The system's performance, usability, and overall value proposition can be further optimized by addressing the identified areas for improvement and incorporating user feedback. Objective 5 To develop a cost-friendly commercialization plan for our product, we undertook comprehensive research into various market adoption strategies and explored potential financial models. Recognizing the importance of collaboration, we initiated and established partnerships with municipalities and other government agencies. These partnerships are intended to facilitate pilot programs and, ultimately, pave the way for broader deployment of our product. Additionally, we recognized the critical role of data collection in our product's functionality. To that end, we dedicated significant effort to identifying cost-effective sensor options that could provide essential data without imposing an undue financial burden on municipalities. Finally, we conducted a thorough cost trade-off analysis. This analysis weighed the potential benefits of hardware investments against the advantages offered by AI-driven analytical tools. The ultimate goal of this analysis was to determine the most efficient and effective approach to data processing and decision-making for our product.?

Publications