SLAP: Shortcut Learning for Abstract Planning
This addresses the problem of limited, manually-defined behaviors in robotic planning, enabling more efficient and generalizable solutions, though it is incremental as it builds on existing TAMP frameworks.
The paper tackled the challenge of long-horizon decision-making in robotics by proposing SLAP, a method that automatically discovers new abstract actions (shortcuts) using reinforcement learning on existing task and motion planning options, resulting in over 50% reduction in plan lengths and higher task success rates compared to baselines.
Long-horizon decision-making with sparse rewards and continuous states and actions remains a fundamental challenge in AI and robotics. Task and motion planning (TAMP) is a model-based framework that addresses this challenge by planning hierarchically with abstract actions (options). These options are manually defined, limiting the agent to behaviors that we as human engineers know how to program (pick, place, move). In this work, we propose Shortcut Learning for Abstract Planning (SLAP), a method that leverages existing TAMP options to automatically discover new ones. Our key idea is to use model-free reinforcement learning (RL) to learn shortcuts in the abstract planning graph induced by the existing options in TAMP. Without any additional assumptions or inputs, shortcut learning leads to shorter solutions than pure planning, and higher task success rates than flat and hierarchical RL. Qualitatively, SLAP discovers dynamic physical improvisations (e.g., slap, wiggle, wipe) that differ significantly from the manually-defined ones. In experiments in four simulated robotic environments, we show that SLAP solves and generalizes to a wide range of tasks, reducing overall plan lengths by over 50% and consistently outperforming planning and RL baselines.