Trust the PRoC3S: Solving Long-Horizon Robotics Problems with LLMs and Constraint Satisfaction

Aidan Curtis, Nishanth Kumar, Jing Cao, Tomás Lozano-Pérez, Leslie Pack Kaelbling

Keywords: LLMs for planning, task and motion planning, constraint satisfaction

TL;DR: having LLM's produce code that corresponds to a CSP, then solving that CSP separately, enables solving long-horizon robotic manipulation tasks with complex and realistic constratins

Abstract: Recent developments in pretrained large language models (LLMs) applied to robotics have demonstrated their capacity for sequencing a set of discrete skills to achieve open-ended goals in simple robotic tasks. In this paper, we examine the topic of LLM planning for a set of *continuously parameterized* skills whose execution must avoid violations of a set of kinematic, geometric, and physical constraints. We prompt the LLM to output code for a function with open parameters, which, together with environmental constraints, can be viewed as a Continuous Constraint Satisfaction Problem (CCSP). This CCSP can be solved through sampling or optimization to find a skill sequence and continuous parameter settings that achieve the goal while avoiding constraint violations. Additionally, we consider cases where the LLM proposes unsatisfiable CCSPs, such as those that are kinematically infeasible, dynamically unstable, or lead to collisions, and re-prompt the LLM to form a new CCSP accordingly. Experiments across simulated and real-world domains demonstrate that our proposed strategy, \OursNoSpace, is capable of solving a wide range of complex manipulation tasks with realistic constraints much more efficiently and effectively than existing baselines.

Website: https://proc3s.csail.mit.edu

Code: https://github.com/Learning-and-Intelligent-Systems/proc3s

Student Paper: yes

Submission Number: 237