Resilient Online Planning for Mobile Robots With Minimal Relaxation of Signal Temporal Logic Specifications

Abstract

We address the problem of resilient motion planning for robots operating under Signal Temporal Logic (STL) specifications in dynamic environments. In such settings, unforeseen events-such as the emergence of dynamic obstacles-can render the original STL specification infeasible. To address this, we propose a reactive framework that enables local corrections or global replanning of the robot's trajectory in response to these unexpected occurrences. When the original STL specification becomes unsatisfiable, our framework involves minimally relaxing it by extending/shrinking time windows or removing some tasks within the user's allowance. This strategy is designed to prevent arbitrarily long delays in mission completion and to facilitate the satisfaction of the mission with minimal temporal relaxation (TR). We present theoretical results supporting our framework and demonstrate its effectiveness through high-fidelity simulations.