Unmanned Aircraft Systems (UAS) have tremendous utility for scientific data collection, delivery, transportation, and military operations. But UAS, like other automated agents, still lack the ability to dynamically refocus their attention -- the kind of self-awareness and dynamic performance adjustment that humans exhibit naturally.
This NSF CAREER project lays the analytical, computing, and tool foundations to make autonomous vehicles capable of assessing and adjusting both performance and resource allocation to maximize holistic system performance. We call this a Cyber-Physical Vehicle Autonomy (CPVA).
Research Thrusts
Mathematical Foundation for Co-regulation
We develop Co-regulated Hybrid Systems (CHS), a new modeling and analysis framework for systems that dynamically vary sampling rate and computational resources to adjust performance. This includes stable-by-construction controllers for multicopters synthesized using the new formalism.
Real-Time Scheduling and Kernel
We design abstract methods, rules, and limits for a cyber resource controller, leading to an open-source Co-regulated Real-Time Kernel (CRTK) that dynamically reallocates computing resources while guaranteeing real-time schedule feasibility.
Cyber-Physical Vehicle Autonomy
We develop co-regulated Markov Decision Processes for online planning and a novel planner pressure metric that combines mission changes, control performance, and environmental uncertainty. The result is an open-source, resource-aware autopilot deployable on platforms like Raspberry Pi, ODROID, Pixhawk, and BeagleBone, with full ROS support.
Evaluation
The CPVA is evaluated in a multi-agent rainforest monitoring scenario where a team of UAS deploy sensors, gather leaf samples, and interact with users -- requiring precise control, learning, perception, and reasoning.
Education and Outreach
The project includes an interdisciplinary CPS education program spanning K-12 through adult populations, including STEM summer camps for middle school girls, senior design capstone projects, REU integration, and public lectures on CPS and automated vehicle safety.