Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Generalized Mission Planning for Heterogeneous Multi-Robot Teams via LLM-constructed Hierarchical Trees

Piyush Gupta, David Isele, Enna Sachdeva, Pin-Hao Huang, Behzad Dariush, Kwonjoon Lee, Sangjae Bae

TL;DR

The paper tackles generalized mission planning for heterogeneous multi-robot teams by introducing an LLM-driven framework that constructs hierarchical task trees to encode complex dependencies and robot capabilities. It combines TAEMS-inspired task representations with a heuristic MRTA decomposition to generate multiple feasible allocations under resource constraints, mediated by predefined subtree routines and function-calling APIs. A $ ho$-bounded MRTA search followed by a topological sort yields precedence-consistent task schedules, while experiments with mobility robots and GPT-4o demonstrate feasible allocations for missions like reuniting a lost child with their mother. The approach enhances flexibility and scalability in real-world settings, though it relies on well-designed subtree routines and may incur suboptimality from greedy pruning and potential LLM hallucinations; future work includes closed-loop simulations and targeted LLM fine-tuning to improve robustness and optimality.

Abstract

We present a novel mission-planning strategy for heterogeneous multi-robot teams, taking into account the specific constraints and capabilities of each robot. Our approach employs hierarchical trees to systematically break down complex missions into manageable sub-tasks. We develop specialized APIs and tools, which are utilized by Large Language Models (LLMs) to efficiently construct these hierarchical trees. Once the hierarchical tree is generated, it is further decomposed to create optimized schedules for each robot, ensuring adherence to their individual constraints and capabilities. We demonstrate the effectiveness of our framework through detailed examples covering a wide range of missions, showcasing its flexibility and scalability.

Generalized Mission Planning for Heterogeneous Multi-Robot Teams via LLM-constructed Hierarchical Trees

TL;DR

The paper tackles generalized mission planning for heterogeneous multi-robot teams by introducing an LLM-driven framework that constructs hierarchical task trees to encode complex dependencies and robot capabilities. It combines TAEMS-inspired task representations with a heuristic MRTA decomposition to generate multiple feasible allocations under resource constraints, mediated by predefined subtree routines and function-calling APIs. A -bounded MRTA search followed by a topological sort yields precedence-consistent task schedules, while experiments with mobility robots and GPT-4o demonstrate feasible allocations for missions like reuniting a lost child with their mother. The approach enhances flexibility and scalability in real-world settings, though it relies on well-designed subtree routines and may incur suboptimality from greedy pruning and potential LLM hallucinations; future work includes closed-loop simulations and targeted LLM fine-tuning to improve robustness and optimality.

Abstract

We present a novel mission-planning strategy for heterogeneous multi-robot teams, taking into account the specific constraints and capabilities of each robot. Our approach employs hierarchical trees to systematically break down complex missions into manageable sub-tasks. We develop specialized APIs and tools, which are utilized by Large Language Models (LLMs) to efficiently construct these hierarchical trees. Once the hierarchical tree is generated, it is further decomposed to create optimized schedules for each robot, ensuring adherence to their individual constraints and capabilities. We demonstrate the effectiveness of our framework through detailed examples covering a wide range of missions, showcasing its flexibility and scalability.

Paper Structure

This paper contains 8 sections, 1 equation, 5 figures, 1 table, 1 algorithm.

Table of Contents

  1. Introduction
  2. Background and Formulation
  3. Mission Planning Framework
  4. Hierarchical Tree construction via LLM function calling
  5. Generating MRTA Alternatives
  6. Results
  7. Limitations and Future Work
  8. Conclusions

Figures (5)

  • Figure 1: Diverse scenarios involving heterogeneous multi-robot team missions. (a) Assisting an old lady cross the street safely, (b) Search and Rescue, (c) Medical emergency, (b) Ensuring safety of a child chasing a ball. These images were created with the assistance of DALL-E 3.
  • Figure 2: An example hierarchical tree. Various subtrees of different color are highlighted in red enclosures. The round nodes represent abstract non-primitive tasks and the rectangular nodes represent executable primitive tasks. Children nodes are connected to their parent node with a logical constraint. The blue arrow represents precedence relationship between tasks.
  • Figure 3: Follow Subtree constructed by the FollowSubtree (agent, resources) routine. Robots that have follow capability in the resources are included in the subtree. The red rectangular nodes represent the primitive actions available to the robots with follow capabilities.
  • Figure 4: Overall mission planning pipeline. LLM utilizes the subtree-routines and APIs to create a hierarchical tree which is decomposed to obtain MTRA alternatives.
  • Figure 5: Hierarchical trees generated by the LLM for different mission goals: (a) Reunite mom with her lost child, (b) Help the woman walking with heavy luggage, (c) Save the city from the monster destroying it, (d) Rescue cat trapped in a building on fire, (e) Recommend best Italian restaurants in the area. The details of the colored subtrees (highlighted with red boxes) used to create the hierarchical tree are omitted for brevity of space.