Robotic and AI systems have strong potential to directly impact our well-being, from self-driving cars to medical robots. Therefore, it is important to consider strong guarantees on the correctness and safety of their behavior. These guarantees ensure the robot will execute the desired behavior and will not execute undesired behavior. The course will define formal notions of system properties such as safety and liveness, explain how to model and analyze those properties in systems that make decisions and act on them, and understand the specific challenges related to making guarantees on embodied AI systems. This course will cover many topics related to formal guarantees of safety and correctness in robotic and AI systems, including temporal logic-based planning, safe control via invariants and control barrier functions, neural net verification, closed loop control with machine learning components, safe reinforcement learning, and other state-ofthe-art topics at the intersection of safety, guarantees, AI, and robotics.
RBE 500
Machine Learning or Intro to AI.