Temporal Logic
Temporal logic (TL) is a formal system for specifying and reasoning about properties that change over time, primarily used to define complex objectives for autonomous systems. Current research focuses on efficiently synthesizing controllers and plans that satisfy TL specifications, often employing techniques like reinforcement learning, large language models (LLMs) for natural language specification translation, and automata-based methods for verification and synthesis. These advancements are significantly impacting robotics, autonomous driving, and other domains requiring safe and reliable control under complex temporal constraints, enabling more robust and adaptable systems.
Papers
Reinforcement Learning with LTL and $ω$-Regular Objectives via Optimality-Preserving Translation to Average Rewards
Xuan-Bach Le, Dominik Wagner, Leon Witzman, Alexander Rabinovich, Luke Ong
Sample-Efficient Reinforcement Learning with Temporal Logic Objectives: Leveraging the Task Specification to Guide Exploration
Yiannis Kantaros, Jun Wang