Advancing optimization-based planning and control of mobile robots under temporal logic specifications

Abstract

The growing adoption of autonomous mobile robots underscores the need for reliable and resilient motion planning and control across diverse applications, ranging from warehouse automation to aerial surveying. Meeting complex, often time-sensitive objectives demands advanced motion planning techniques. In this regard, temporal logics, such as Signal Temporal Logic (STL), serve as rigorous and compact frameworks for encoding mission specifications, encompassing logical, spatial, and temporal constraints. However, designing control algorithms that fully satisfy these specifications—particularly in scenarios requiring multi-agent collaboration, resilience to unforeseen events, and real-time decision-making—presents significant challenges. This dissertation tackles these challenges through three primary objectives. First, it investigates scalable frameworks for multi-agent systems to fulfill collective temporal logic specifications, including preemptable tasks agents can complete asynchronously. It combines sampling-based trajectory generation (e.g., RRT*) with mixed-integer programming (MIP), to address missions involving heterogeneous agents. Second, it explores resilient motion planning strategies, introducing a quantitative metric to minimize specification violations when constraints are breached. This metric captures cumulative task relaxation, allowing structural adjustments in STL specifications, such as modifying task time intervals or removing tasks when necessary. Lastly, the dissertation addresses real-time STL-based planning for missions requiring fast decision-making, such as completing tasks defined on noncooperative targets, by constructing control barrier functions (CBFs) based on the robots’ actuation limits and an optimized sequence of STL tasks. The proposed frameworks and algorithms are validated through theoretical analyses, simulations, and experiments, demonstrating their scalability and effectiveness in complex robotic applications. Through these contributions, this work advances temporal logic-based control, enhancing STL’s applicability to resilient and real-time motion planning. It provides essential insights and tools for mobile robots capable of autonomous operation in evolving environments with intricate mission requirements.