Context-Triggered MPC Lets Robots Handle Complex Logical Tasks Safely

Arash Bahari Kordabad and coauthors propose a context-triggered robust MPC framework that ensures discrete-time linear systems satisfy context-dependent linear temporal logic specifications despite additive bounded disturbances. Their switching control architecture combines robust MPC with a local invariant controller, where the MPC value function serves as a reachability certificate, avoidance is enforced via robust constraints, and the stay condition is enforced by the local controller. Convex duality reformulates the robust constraints into equivalent deterministic quadratic and second-order cone programs. Demonstrated on a robot navigation problem with static and moving environments, the method yields significantly larger feasible sets than Lyapunov-based approaches while naturally accommodating dynamic environments and online task reconfiguration.