Safe Control and Learning Using Generalized Action Governor
This work addresses safety-critical control and learning for systems with constraints, offering a supervisory scheme that ensures constraint satisfaction during operation and learning, which is incremental as it builds on existing action governor concepts.
The paper tackles the problem of enforcing state and input constraints in control systems under uncertainties by introducing the Generalized Action Governor, which adjusts actions online to maintain safety. It presents design procedures for linear and discrete systems and integrates the governor with safe learning strategies like Q-learning and Koopman-based control, demonstrating effectiveness through numerical results.
This paper introduces the Generalized Action Governor (AG), a supervisory scheme that augments a nominal closed-loop system with the capability to enforce state and input constraints through online action adjustment. We develop a generalized AG theory for discrete-time systems under bounded uncertainties, and relax the usual requirement of positive invariance to returnability of a safe set. Based on the theory, we present tailored AG design procedures for linear systems and for discrete systems with finite state and action spaces. We further study safe online learning enabled by the AG and present two safe learning strategies, namely safe Q-learning and safe data-driven Koopman operator-based control, both integrated with the AG to guarantee constraint satisfaction during learning. Numerical results illustrate the proposed methods.