Multi-Robot Learning-Informed Task Planning under Uncertainty
Abhish Khanal, Abhishek Paudel, Hung Pham, Gregory Stein
AI summary
Problem
Coordinating multi-robot teams to complete complex tasks in partially known environments is challenging due to uncertain object locations, concurrent durative actions, and combinatorial outcome growth over long planning horizons.
Approach
The authors combine a learned estimator for action-outcome probabilities with a model-based planning framework that uses task progress automata and partially observable Monte Carlo tree search to coordinate robot actions until task completion.
Key results
- Up to 47.0% cost reduction over baselines for 1-robot teams in ProcTHOR simulations
- Up to 40.7% improvement for 2-robot teams and 33.8% for 3-robot teams
- Successful real-world deployment with a two-robot LoCoBot team in household settings
- Novel belief-state advancement algorithm that efficiently tracks concurrent robot progress and search outcomes
Why it matters
Provides a scalable, learning-augmented planning framework that enables real-world multi-robot teams to efficiently complete complex search and manipulation tasks in unpredictable environments.
Abstract
We want a multi-robot team to complete complex tasks in minimum time where the locations of task-relevant objects are not known. Effective task completion requires reasoning over long horizons about the likely locations of task- relevant objects, how individual actions contribute to overall progress, and how to coordinate team efforts. Planning in this setting is extremely challenging: even when task-relevant infor- mation is partially known, coordinating which robot performs which action and when is difficult, and uncertainty introduces a multiplicity of possible outcomes for each action, which further complicates long-horizon decision-making and coordination. To address this, we propose a multi-robot planning abstraction that integrates learning to estimate uncertain aspects of the environ- ment with model-based planning for long-horizon coordination. We demonstrate the efficient multi-stage task planning of our approach for 1, 2, and 3 robot teams over competitive baselines in large ProcTHOR household environments. Additionally, we demonstrate the effectiveness of our approach with a team of two LoCoBot mobile robots in real household settings.