LTLCodeGen: Code Generation of Syntactically Correct Temporal Logic for Robot Task Planning

2025-03-10

AI Summary

This work addresses the semantic gap between natural-language navigation instructions and robotic planning by proposing an LLM-driven, code-generation-based method for translating colloquial directives into syntactically correct and semantically grounded Linear Temporal Logic (LTL) formulas. The approach integrates semantic occupancy maps with a modular task-planning framework to enable end-to-end, verifiable, collision-free path generation from speech input. Key contributions are: (i) the first application of code-generation paradigms to LTL formula synthesis, leveraging structured prompt engineering to ensure syntactic validity and logical consistency; and (ii) enabling interpretable, formally verifiable navigation behaviors. Experiments in simulation and on real robotic platforms demonstrate substantial improvements over end-to-end LLM planners and existing LLM-to-LTL baselines, with strong robustness and generalization across diverse instruction styles and environments.

Technology Category

• Planning, Routing, and Scheduling: Planning with Language Models
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Humans and AI: Human-Aware Planning and Behavior Prediction

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

This paper focuses on planning robot navigation tasks from natural language specifications. We develop a modular approach, where a large language model (LLM) translates the natural language instructions into a linear temporal logic (LTL) formula with propositions defined by object classes in a semantic occupancy map. The LTL formula and the semantic occupancy map are provided to a motion planning algorithm to generate a collision-free robot path that satisfies the natural language instructions. Our main contribution is LTLCodeGen, a method to translate natural language to syntactically correct LTL using code generation. We demonstrate the complete task planning method in real-world experiments involving human speech to provide navigation instructions to a mobile robot. We also thoroughly evaluate our approach in simulated and real-world experiments in comparison to end-to-end LLM task planning and state-of-the-art LLM-to-LTL translation methods.

Problem

• Translate natural language to correct LTL for robot tasks.
• Generate collision-free robot paths from semantic maps.
• Compare LLM-based task planning with existing methods.

Innovation

• LLM translates natural language to LTL
• Semantic occupancy map guides motion planning
• LTLCodeGen ensures syntactically correct LTL generation