Learning to Satisfy Unknown Constraints in Iterative MPC
This addresses the challenge of ensuring safety in control systems when constraints are unknown, which is crucial for applications like robotics and autonomous systems, though it is incremental as it builds on existing MPC and learning methods.
The paper tackles the problem of controlling linear time-invariant systems with unknown polyhedral state constraints by iteratively learning these constraints from closed-loop trajectory data and designing an MPC controller to satisfy them robustly, providing robust and probabilistic guarantees of constraint satisfaction based on the number of task iterations.
We propose a control design method for linear time-invariant systems that iteratively learns to satisfy unknown polyhedral state constraints. At each iteration of a repetitive task, the method constructs an estimate of the unknown environment constraints using collected closed-loop trajectory data. This estimated constraint set is improved iteratively upon collection of additional data. An MPC controller is then designed to robustly satisfy the estimated constraint set. This paper presents the details of the proposed approach, and provides robust and probabilistic guarantees of constraint satisfaction as a function of the number of executed task iterations. We demonstrate the safety of the proposed framework and explore the safety vs. performance trade-off in a detailed numerical example.