Multi-Robot Exploration with Inferred Human Intent in Large-Scale Field Experiments: Lessons Learnt

Published

14-10-2024

Type of Work

Article

Abstract

We provide lessons learnt from developing a multi-robot system (MRS) for informative path planning in a large-scale field environment. Our heterogeneous MRS explores a hazardous area to ensure the safety of a human. Robots prioritise exploration of where the human will travel from a high-level goal without relying on constant human input. We combined Decentralised Monte Carlo Tree Search with a Human Intent Inference engine to enable robots to explore ahead of a human, prioritising exploration in unseen areas that the human is predicted to travel through. When hazards are identified, they are communicated to the human. Field experiments with three robots were conducted in an 800 m × 400 m field environment. Robots successfully searched areas the human travelled through in advance to detect hazards, navigating diverse terrain with resilient algorithms. The primary lessons learnt concern robust algorithm design for outdoor MRSs, and the challenges posed by developmental robotic platforms.

Citation

Bray, E., Brown, M., Lee, K. M. B., Linstrom, F., Tan, A., Eyers, A., ... & Best, G. (2024). Multi-Robot Exploration with Inferred Human Intent in Large-Scale Field Experiments: Lessons Learnt Conference, IROS,  Australia. https://www.researchgate.net/publication/384893740_Multi-Robot_Exploration_with_Inferred_Human_Intent_in_Large-Scale_Field_Experiments_Lessons_Learnt

Share

COinS