Coding Agents for Generalized Task and Motion Planning Problems

📄 arXiv: 2609.30233v1 📥 PDF

作者: Matteo Merler, Bowen Li, Josh Roy, Yichao Liang, Qianwei Wang, Yixuan Huang, Tom Silver

分类: cs.RO, cs.AI

发布日期: 2026-09-24

备注: 9 pages, 4 figures, 3 tables


💡 一句话要点

提出编码代理以解决广义任务与运动规划问题

🎯 匹配领域: 支柱一:机器人控制 (Robot Control) 支柱九:具身大模型 (Embodied Foundation Models)

关键词: 任务与运动规划 编码代理 程序合成 自动化 智能系统

📋 核心要点

  1. 现有的任务与运动规划方法在处理复杂约束时效率低下,且需要大量的手工工程。
  2. 本文提出通过编码代理自动化程序合成,利用环境交互来生成可泛化的规划程序。
  3. 实验结果显示,编码代理在多种环境中成功率高达95%,显著优于传统手工规划方法。

📝 摘要(中文)

任务与运动规划(TAMP)问题即使在完全可观察和以对象为中心的状态下仍然困难,因为离散决策与几何、运动学和动态约束紧密耦合。广义TAMP通过利用问题实例之间的规律性来减少新实例的规划工作量。然而,现有方法需要大量特定于TAMP的工程。本文探讨编码代理是否能够通过合成跨实例的程序来自动化这一过程。给定任务描述和模拟器访问,每个代理选择如何与环境互动,同时在固定的合成预算内开发程序。我们在28个来自KinDER和PDDLStream的模拟环境中评估Claude Code和Codex,发现编码代理在广义TAMP中表现出色,成功率显著高于手工设计的规划器。

🔬 方法详解

问题定义:本文旨在解决任务与运动规划(TAMP)中的复杂性问题,现有方法在处理几何和动态约束时效率低下,且需要大量手工工程。

核心思路:通过编码代理自动合成程序,利用环境交互来生成可泛化的规划程序,从而减少对手工设计的依赖。

技术框架:整体架构包括任务描述输入、环境交互模块和程序合成模块。代理在固定预算内与环境互动,生成并评估程序。

关键创新:最重要的创新在于使用编码代理进行程序合成,能够在多个实例中实现泛化,显著降低了计算复杂度。

关键设计:在程序合成过程中,代理通过与环境的互动来校准物理模型、测试边界情况和优化策略,确保生成的程序在未见实例上也能有效运行。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

实验结果表明,编码代理在16个环境中平均成功率达到95%,而手工设计的规划器仅为47%。在对象数量增加的情况下,代理的程序成功率保持更高,且每个实例的计算量减少了一个数量级。

🎯 应用场景

该研究的潜在应用领域包括机器人导航、自动化制造和智能家居等场景。通过提高任务与运动规划的效率,能够在复杂环境中实现更高效的自动化操作,未来可能推动智能系统的广泛应用。

📄 摘要(原文)

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.