Coding Agents for Generalized Task and Motion Planning Problems

2026-09-24

AI Summary

This study addresses the engineering complexity arising from the coupling of discrete decision-making and physical constraints in generalized Task and Motion Planning (TAMP). To this end, it proposes an automated program synthesis framework based on coding agents. By leveraging advanced coding agents to interact with environments and optimize strategies, the method automatically generates programs that generalize across instances, eliminating the need for manual engineering tailored to specific TAMP scenarios. Experimental results demonstrate that the proposed approach achieves average success rates ranging from 56% to 95% while reducing computational overhead by an order of magnitude. Furthermore, its generalization capability significantly surpasses that of traditional handcrafted planners. All source code has been made publicly available.

Abstract

Task and motion planning (TAMP) problems remain difficult even with full observability and object-centric states because discrete decisions are tightly coupled to geometric, kinematic, and dynamic constraints. Generalized TAMP addresses this difficulty by exploiting regularities across problem instances to reduce planning effort on new instances. However, existing methods require substantial TAMP-specific engineering. We investigate whether coding agents can automate this process by synthesizing programs that generalize across instances. Given a task description and simulator access, each agent chooses how to interact with the environment while developing a program within a fixed synthesis budget. The program is then frozen and evaluated on unseen instances. We evaluate Claude Code (Opus 5) and Codex (GPT-5.6 Sol and GPT-6 Astra) on 28 simulated environments from KinDER and PDDLStream, with object counts beyond those evaluated in the original benchmark. Across all program synthesis methods, we evaluate 980 generated programs on 100 held-out instances each, 98,000 evaluation episodes in total. Overall, we find that coding agents are surprisingly effective at generalized TAMP: all three agent configurations outperform hand-engineered planners, one-shot generation, and an LLM-based generalized planning baseline in mean success (56% to 95% versus 47% for the planners, on the 16 environments where a planner is available). As object counts grow, the agents' programs maintain higher success than the planner, using an order of magnitude less computation per instance on average. Logs show agents using interaction to calibrate physical models, test edge cases, and refine strategies. We release all code, including the full prompts given to the agents. These findings suggest that coding agents are a strong baseline for generalized TAMP.

Problem

• Task and Motion Planning
• Generalized TAMP
• Coding Agents
• Program Synthesis
• Generalization

Innovation

• Coding Agents
• Generalized Task and Motion Planning
• Program Synthesis
• Large Language Models
• Simulator Interaction