Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Temporal Planning Framework for Multi-Agent Systems via LLM-Aided Knowledge Base Management

Enrico Saccon, Ahmet Tikna, Davide De Martini, Edoardo Lamon, Luigi Palopoli, Marco Roveri

TL;DR

This paper addresses the challenge of converting natural-language task descriptions into reliable, explainable plans for multi-agent robotics. It introduces PLANTOR, a framework that couples LLM-assisted Prolog knowledge-base generation with a three-stage planning pipeline (TO planning, PO planning with enablers and resources, and MILP-based optimization) and finally maps plans to Behavior Trees for ROS2 execution. The key contributions include a two-phase KB construction (high-level and low-level) aided by chain-of-thought prompting, a formal mapped temporal planning formalism, and an MILP-based scheduler that enables parallel execution while respecting resource constraints. Experimental validation in block-world and arch-building scenarios, plus a real-world dual-robot arch construction, demonstrates the approach’s feasibility, explainability, and potential practicality while highlighting limitations in scalability and LLM reliability, guiding future work toward tighter integration with state-of-the-art planners and motion planning. Overall, PLANTOR advances explainable, scalable, multi-agent planning by effectively blending symbolic reasoning with data-driven KB generation and optimization, enabling robust ROS2 deployment in real-world robotics tasks.

Abstract

This paper presents a novel framework, called PLANTOR (PLanning with Natural language for Task-Oriented Robots), that integrates Large Language Models (LLMs) with Prolog-based knowledge management and planning for multi-robot tasks. The system employs a two-phase generation of a robot-oriented knowledge base, ensuring reusability and compositional reasoning, as well as a three-step planning procedure that handles temporal dependencies, resource constraints, and parallel task execution via mixed-integer linear programming. The final plan is converted into a Behaviour Tree for direct use in ROS2. We tested the framework in multi-robot assembly tasks within a block world and an arch-building scenario. Results demonstrate that LLMs can produce accurate knowledge bases with modest human feedback, while Prolog guarantees formal correctness and explainability. This approach underscores the potential of LLM integration for advanced robotics tasks requiring flexible, scalable, and human-understandable planning.

A Temporal Planning Framework for Multi-Agent Systems via LLM-Aided Knowledge Base Management

TL;DR

This paper addresses the challenge of converting natural-language task descriptions into reliable, explainable plans for multi-agent robotics. It introduces PLANTOR, a framework that couples LLM-assisted Prolog knowledge-base generation with a three-stage planning pipeline (TO planning, PO planning with enablers and resources, and MILP-based optimization) and finally maps plans to Behavior Trees for ROS2 execution. The key contributions include a two-phase KB construction (high-level and low-level) aided by chain-of-thought prompting, a formal mapped temporal planning formalism, and an MILP-based scheduler that enables parallel execution while respecting resource constraints. Experimental validation in block-world and arch-building scenarios, plus a real-world dual-robot arch construction, demonstrates the approach’s feasibility, explainability, and potential practicality while highlighting limitations in scalability and LLM reliability, guiding future work toward tighter integration with state-of-the-art planners and motion planning. Overall, PLANTOR advances explainable, scalable, multi-agent planning by effectively blending symbolic reasoning with data-driven KB generation and optimization, enabling robust ROS2 deployment in real-world robotics tasks.

Abstract

This paper presents a novel framework, called PLANTOR (PLanning with Natural language for Task-Oriented Robots), that integrates Large Language Models (LLMs) with Prolog-based knowledge management and planning for multi-robot tasks. The system employs a two-phase generation of a robot-oriented knowledge base, ensuring reusability and compositional reasoning, as well as a three-step planning procedure that handles temporal dependencies, resource constraints, and parallel task execution via mixed-integer linear programming. The final plan is converted into a Behaviour Tree for direct use in ROS2. We tested the framework in multi-robot assembly tasks within a block world and an arch-building scenario. Results demonstrate that LLMs can produce accurate knowledge bases with modest human feedback, while Prolog guarantees formal correctness and explainability. This approach underscores the potential of LLM integration for advanced robotics tasks requiring flexible, scalable, and human-understandable planning.

Paper Structure

This paper contains 42 sections, 10 equations, 4 figures, 6 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Problem Description and Solution Overview
  3. Problem Formalisation
  4. Solution Overview
  5. Background and Definitions
  6. Task Planning
  7. Classical Task Planning
  8. Temporal task planning
  9. Mapped Temporal Task Planning
  10. Prolog
  11. Large Language Models
  12. Knowledge-base Generation
  13. Runing Example -- KMS
  14. Plan Generation
  15. Total-Order Plan Generation
  16. ...and 27 more sections

Figures (4)

  • Figure 1: The architecture of the proposed framework.
  • Figure 2: A scheme showing the running example. Two blocks must be moved from their initial position to a new position in which they are also stacked.
  • Figure 3: Instance of the examples fed to the LLM through few-shot prompting.
  • Figure 4: The real-life experiment we carried out using two Universal Robots arms.