One of the grand technical challenges of our generation is to get ready to feed 9 billion people by 2050 with sustainable use of water and chemicals. However, we are facing unprecedented challenges in adopting sustainable agricultural management practices, increasing production, keeping agriculture profitable and coping with deadly biotic and abiotic stresses and diseases as well as changing climate that threaten yield. This project aims to transform Cyber-Physical System (CPS) capabilities in agriculture to enable farmers to respond to crop stressors with lower cost, greater agility, and significantly lower environmental impact than current practices. The objective is to make foundational advances in AI, machine learning and robotics to individual plant-level sensing, modeling and reasoning. This enables small autonomous dexterous robots instead of the heavy farm equipment to monitor plants or small plots individually and treat them with minimum amount of chemicals. This also lowers the barrier to entry for small scale farmers, increases safety, minimizes runoff as well as soil compaction. This project includes a significant collaboration with the University of Illinois at Urbana-Champaign that is funded by the National Institute of Food and Agriculture (NIFA) within the U.S. Department of Agriculture.
The availability of vast amounts of operational and end-user data in cyber-physical systems implies that paradigm improvements in monitoring and control can be attained via learning by many artificial intelligence agents despite them possessing vastly different abilities. Engaging this heterogeneous agent base in the context of the smart grid requires the use of hierarchical markets, wherein end-users participate in downstream markets collectively through aggregators, who in turn are coordinated by an upstream market. The goal of this project is to conduct a systematic study of such market-mediated learning and control. This project aims at much deeper levels of participation from end-users contributing electricity generation such as rooftop solar, shedding load via demand response, and providing storage capabilities such as electric vehicle batteries, to transform into reliable distributed energy resources (DER) at the level of wholesale markets. A methodological theme is multi-agent reinforcement learning (MARL) by agents that control physical systems via actions at different levels of the hierarchy. Underlying the whole project are well-founded physical models of the transmission and distribution grids, which provide structure to the problem domain and concrete use cases. This project facilitates a deeper level of decarbonization in the electricity sector, and contributes to climate change solutions by engineering a flat, interactive grid architecture that allows significant DERs to provide electricity services to both local and regional grids. Engagement with a grid-level market operator enables the project to address a problem space of immediate relevance to the current electricity grid. The project also includes the development of educational materials on data-analytics and energy systems. Intrinsic to the program are efforts at outreach to involve high-school students via demonstrations and lectures based on the technology developed.
Understanding complex activity due to humans and vehicles in a large environment like a city neighborhood or even an entire city is one of the main goals of smart cities. The activities are heterogeneous, distributed, vary over time and mutually interact in many ways, making them hard to capture and understand and mitigate issues in a timely manner. While there has been tremendous progress in capturing aggregate statistics that helps in traffic and city management as well as personal planning and scheduling, much of this work ignores anomalous patterns. Examples include protests, erratic driving, near accidents, construction zone activity, and numerous others. Discovering and resolving anomalies is challenging for many reasons as they are complex and rare, depend on the context and depend on the spatial and temporal extent over which they are observed. There are potentially a large number of anomalies or anomalous patterns, so they are impossible to label and describe manually.
This project investigates fundamental techniques for building mathematical models that can be safely used to make trustworthy predictions and control decisions. Mathematical models form the foundation for modern Cyber-Physical Systems (CPS). Examples include vehicle models that predict how a car will move when brakes are applied, or physiological models that predict how the blood glucose levels change in a patient with type-1 diabetes when insulin is administered. The success of machine learning tools has yielded data-driven models such as neural networks. However, depending on how data is collected and the models are learned, it is possible to obtain models that violate fundamental physical, chemical, or physiological facts that can potentially threaten life and property. The approach of the project is to expose these model flaws through advanced analysis. The project seeks to broaden participation in computing through mentoring activities that will encourage undergraduate women and members of underrepresented minority groups to consider a career in research.
Advances in artificial intelligence (AI) implemented with neural networks and other machine learning techniques have transformed what computers can accomplish. Despite their potential, AI has had comparatively less impact on cyber-physical systems (CPS). Many CPS interact with the physical world where safety is important, so a solution with superior performance 99.9% of the time may still be unacceptable for a CPS. Unfortunately, AI systems are hard to prove correct ? it is difficult to trust the systems will always do what they are designed to do. The goal of this research is to advance the foundations of formal methods in order to make formal verification of AI-based CPS practical. If successful, the work will enable a justified trust in AI systems and allow AI to be applied within safety-critical processes that interact with the physical world. The project investigates approximation approaches where an AI component is replaced by an approximation with similar performance that is easier to verify.
The goal of this project is to achieve coordination and localization among robots, even if some of the robots are behaving in an untrustworthy way. The approach is to use communication signals, and to control the motion of some robots, to learn about the environment and other agents in a way that provably supports coordinated behaviors. Multi-agent Cyber-Physical Systems (CPS) are poised for impact in society as self-driving cars, delivery drones, and disaster response robots. The results from this project will lead to improved mission intelligence for robotic platforms with significance across many areas: from Search and Rescue (SAR) tasks, to CubeSats and space exploration. The project includes tight integration between research tasks and educational activities, including participation in Hack for Humanity where students will explore SAR tasks.
The aim of this proposal is to enable people to control robots remotely using virtual reality. Using cameras mounted on the robot and a virtual reality headset, a person can see the environment around the robot. However, controlling the robot using existing technologies is hard: there is a time delay because it?s slow to send high quality video over the Internet. In addition, the fidelity of the image is worse than looking through human eyes, with a fixed and narrow view. This proposal will address these limitations by creating a new system which understands the geometry and appearance of the robot's environment. Instead of sending high-quality video over the Internet, this new system will only send a smaller amount of information about how the environment's geometry and appearance has changed over time. Further, understanding the geometry and appearance will let us expand the view visible to the person. Overall, these will improve a human's ability to remotely control the robot by increasing fidelity and responsiveness. We will demonstrate this technology on household tasks, on assembly tasks, and by manipulating small objects.
Most preK-12 school districts in the United States dedicate significant resources to safeguard against active shooters, e.g., school hardening, community planning, identification of suspicious behavior, crisis training for law enforcement, and training exercises for students, teachers, and all school personnel. However, when such an active-shooting event is in progress, only vague guidance is available to students and school personnel in the form of directives such as the "run-hide-fight" protocol. The Active Shooter Tracking and Evacuation Routing for Survival (ASTERS) project will complement these efforts by tracking a shooter in real time across multiple cameras and microphones, calculate the optimum evacuation path to safety for each student, teacher, and staff member, and communicate this information through a mobile app interface that is co-created in partnership with a connected community of students, parents, educators and administrators as well as school resource officers and school safety officers. ASTERS will incorporate multi-modal sensing, machine learning and signal processing techniques to accurately localize a gunman and weapons while preserving privacy of school community members. It will also use new computer vision and high-performance computing solutions to estimate crowd density and movement of people, and novel optimization and real-time simulation algorithms to predict ideal evacuation routes based on the building layout and predicted movement of the shooter. ASTERS will collaborate with schools to develop an annotated, multi-modal active shooter data set using a combination of digital simulation data and real-life practice drills. The research team will also partner with first-responders to ensure that ASTERS aligns with their needs.
The availability of vast amounts of operational and end-user data in cyber-physical systems implies that paradigm improvements in monitoring and control can be attained via learning by many artificial intelligence agents despite them possessing vastly different abilities. Engaging this heterogeneous agent base in the context of the smart grid requires the use of hierarchical markets, wherein end-users participate in downstream markets collectively through aggregators, who in turn are coordinated by an upstream market. The goal of this project is to conduct a systematic study of such market-mediated learning and control. This project aims at much deeper levels of participation from end-users contributing electricity generation such as rooftop solar, shedding load via demand response, and providing storage capabilities such as electric vehicle batteries, to transform into reliable distributed energy resources (DER) at the level of wholesale markets. A methodological theme is multi-agent reinforcement learning (MARL) by agents that control physical systems via actions at different levels of the hierarchy. Underlying the whole project are well-founded physical models of the transmission and distribution grids, which provide structure to the problem domain and concrete use cases. This project facilitates a deeper level of decarbonization in the electricity sector, and contributes to climate change solutions by engineering a flat, interactive grid architecture that allows significant DERs to provide electricity services to both local and regional grids. Engagement with a grid-level market operator enables the project to address a problem space of immediate relevance to the current electricity grid. The project also includes the development of educational materials on data-analytics and energy systems. Intrinsic to the program are efforts at outreach to involve high-school students via demonstrations and lectures based on the technology developed.
Across the country, dairy farmers are extremely vulnerable to a variety of factors, including cattle health and wellness. For example, when a cow is ill or stressed, she produces less milk. Subsequently, she might develop secondary conditions because she doesn't want to compete for food or water. For example, if a cow has separated herself from a herd, this may indicate that she is sick. Modern farmers already use a great deal of biometric information to monitor their animals, but social and psychological factors of cattle have not been well-studied. The purpose of the proposed project is to build a cyber-physical system that integrates the social interactions of dairy cattle with other biometric data, develop predictive models that use such data to perform early identification of sick or vulnerable cattle, and create algorithms to provide adaptive interventions to the cattle farmers. This project has the potential to lead to substantial impacts, both scientific and commercial. The decline of small farms is a well-known problem across the country, and anything that helps improve farmers? profit margins is valuable. This project will result in general guidelines about herd management- which could act as a ?tutorial? of sorts to new farmers- as well as technologies for individual herd management.
The proposed project will be one of the first to study the interaction networks of domestic cattle herds and attempt to tie those networks to biometric data and illness. On the sensing side, major contributions will include merging location/interaction data with other biometric data to detect social behaviors. On the analysis and action side, contributions will include connecting network behaviors of dairy cattle to their health and wellness, learning recommendation rules from farmers? responses on different cattle management scenarios through inductive logic programming, and designing explainable algorithms to provide recommendations for addressing health and wellness problems. Experimentation will be conducted using selected dairy farms in Northeast USA.