Robust MITL planning under uncertain navigation times

• DOI: 10.48550/arxiv.2403.03727
• Publication Date: 2024-03-06
• Authors: Alexis Linard; Anna Gautier; Daniel Duberg; Jana Tůmová

In environments like offices, the duration of a robot's navigation between two locations may vary over time. For instance, reaching kitchen take more time during lunchtime since corridors are crowded with people heading same way. this work, we address problem routing in such tasks expressed Metric Interval Temporal Logic (MITL) - rich robot task specification language that allows us to capture explicit requirements. Our objective is find strategy maximizes temporal robustness MITL task. As first step towards solution, define Mixed-integer linear programming approach solving planning Varying Weighted Transition System, where durations deterministic but depending on day. Then, apply planner optimize for Markov Decision Processes, physical uncertain, time-dependent distribution possible delays known. Finally, develop receding horizon Processes preserves guarantees robustness. We show scalability our algorithms simulations robotic tasks.