Events

Learning and Control for Safety, Efficiency, and Resiliency of Cyber-Physical Systems

The rapid evolution of ubiquitous sensing, communication, and computation technologies has contributed to the revolution of cyber-physical systems (CPS). Learning-based methodologies are integrated to the control of physical systems and provide tremendous opportunities for AI-enabled CPS. However, existing networked CPS decision-making frameworks lack understanding of the tridirectional relationship among communication, learning and control. It remains challenging to leverage the communication capability for the learning and control methodology design of CPS, to improve the safety, efficiency, and robustness of the system. In the first part of the talk, we present our research contributions on learning and control with information sharing for networked CPS. We design the first uncertainty quantification method for collaborative perception of connected autonomous vehicles (CAVs), and show the accuracy improvement and uncertainty reduction performance of our method. To utilize the information shared among agents, we then develop a safe and scalable deep multi-agent reinforcement learning (MARL) algorithms to improve system safety and efficiency. We validate the benefits of communication in MARL especially for CAVs under challenging mixed traffic scenarios. To motivate agents to communicate and coordinate, we design a novel stable and efficient Shapley value-based reward reallocation scheme for MARL. Finally, considering the complicated system dynamics and state information uncertainties from sensors and learning-based perception of networked CPS, we present our contribution to robust MARL methods, including formal analysis on the solution concept of MARL under state uncertainties and state perturbations. In the second part of the talk, we briefly present our research contributions on autonomous mobility-on-demand (AMoD) systems for fair, efficient, and sustainable transportation, with data-driven robust optimization theorems, algorithm design and real-world data validations. We will also highlight our research results on CPS security and our future work to bridge the gap between theories and real-world applications for CPS.