Scalable Task Planning via Large Language Models and Structured World Representations

2024-09-07

AI Summary

To address the computational intractability of task-level planning in large-scale environments—caused by state-space explosion—this paper proposes an LLM-driven symbolic-neural hybrid planning framework. Our core innovation is the first use of a large language model as a lightweight semantic filter, jointly leveraging structured world modeling and commonsense reasoning to enable semantic-aware state-space pruning. This approach preserves the interpretability of symbolic planning while substantially improving generalization and computational efficiency: planning time is reduced by 62% in a household simulation environment; and multi-step real-world manipulation tasks are successfully executed on a 7-DOF robotic arm, demonstrating practicality and robustness in complex physical settings. The framework bridges high-level semantic reasoning with low-level control, enabling scalable and interpretable task planning without sacrificing execution fidelity.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Planning, Routing, and Scheduling: Planning with Language Models
• Multiagent Systems: Multiagent Planning

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Planning methods struggle with computational intractability in solving task-level problems in large-scale environments. This work explores leveraging the commonsense knowledge encoded in LLMs to empower planning techniques to deal with these complex scenarios. We achieve this by efficiently using LLMs to prune irrelevant components from the planning problem's state space, substantially simplifying its complexity. We demonstrate the efficacy of this system through extensive experiments within a household simulation environment, alongside real-world validation using a 7-DoF manipulator (video https://youtu.be/6ro2UOtOQS4).

Problem

• Scalable task planning in large environments
• Leveraging LLMs for commonsense knowledge
• Pruning state space to simplify complexity

Innovation

• LLMs prune state space
• Structured world representations simplify
• Household simulation validates efficacy