Reaction Templates: A Formal Approach to Realize Reactivity in Task and Motion Planning-Based Action Execution
Anne Köpken, Adrian Simon Bauer, Nesrine Batti, Daniel Leidner
AI summary
Problem
Robots struggle to execute prolonged, robust action sequences autonomously due to limitations in handling unforeseen situations or failures during execution. Existing planning frameworks often intertwine error recovery with action definitions, causing code repetition, high complexity, and poor scalability.
Approach
Introduces Reaction Templates (RTs), a formal framework that runs concurrently with task execution to automatically monitor violated assumptions and trigger customizable recovery strategies. This design separates planned actions from reactive behaviors while allowing hierarchical parameter tuning for different contexts.
Key results
- Novel paradigm separating action strategies from reaction strategies
- Formal definition of Reaction Templates with activation, monitoring, and reaction sections
- Hierarchical parameter system for customizing reactions to specific actions and objects
- Proof-of-concept implementation and experimental validation on the Rollin’ Justin humanoid robot
Why it matters
Enables robots to robustly execute complex, long-horizon manipulation tasks in dynamic real-world environments by providing a scalable, reusable framework for real-time failure detection and recovery.
Abstract
Recent advancements in artificial intelligence have broadened the spectrum of tasks that robots can effectively tackle. However, the seamless execution of prolonged action sequences continues to pose a considerable challenge, at- tributed to limitations in the abilities of today’s planning- based robots to react to unforeseen situations or failures. In response, we introduce Reaction Templates (RTs), a for- mal approach for integrating reactivity into task and motion planning. Operating concurrently with the primary execution logic, RTs enable a clear differentiation between planned actions and the necessary recovery strategies for handling unexpected events. This design promotes scalability by es- tablishing reusable building blocks and customizable parame- ters, thereby enhancing flexibility in application. We provide a thorough introduction to the RT concept, elucidating its principles, mechanisms, and the rationale behind its design decisions. The resulting benefits of the approach are demon- strated through experimental validation with the humanoid robot Rollin’ Justin. A video highlighting the core concepts of the paper and depicting the experimental validation can be viewed at https://www.youtube.com/watch?v=aiLKXQCcMy4.