Multi-agent reinforcement curriculum learning for real unmanned ground vehicles

Timothy Mead; Zhe Wang; Ernest Foo; Jin Song Dong; Naipeng Dong; Ryan Ko; Abigail Koay; Kien Nguyen; Yue Xu; Junae Kim; Stephen Bornstein; Zhe Hou

doi:10.1016/j.engappai.2026.113780

Back

Multi-agent reinforcement curriculum learning for real unmanned ground vehicles

Journal article

Open access

Peer reviewed

Multi-agent reinforcement curriculum learning for real unmanned ground vehicles

Timothy Mead, Zhe Wang, Ernest Foo, Jin Song Dong, Naipeng Dong, Ryan Ko, Abigail Koay, Kien Nguyen, Yue Xu, Junae Kim, …

Engineering Applications of Artificial Intelligence, Vol.167, pp.1-13

2026

DOI: https://doi.org/10.1016/j.engappai.2026.113780

Files and links (1)

pdf

9912062785026213.00 MBDownload View

Published VersionCC BY V4.0, Open Access

Abstract

Autonomous systems

Curriculum learning

Federated learning

Multi-agent

Reinforcement learning

Robotics

Swarm

Unmanned ground vehicles

This paper investigates the use of deep reinforcement learning (DRL) for the control of mobile robot teams within the context of navigation and task-based collaborative scenarios. We apply a DRL policy with a tailored neural network architecture as a solution to control, path planning, and higher-level guidance tasks. Our network architecture was trained using a unique multi-stage curriculum that progresses from single-agent navigation, to multi-agent pathfinding with obstacles, and finally to a complex collaborative firefighting scenario. This structured approach accelerates training convergence by systematically building sophisticated collaborative behaviours upon foundational skills, which enhances training stability and guides the agents towards learning effective and coordinated strategies The policy evaluation was conducted in both simulation and hybrid simulation-physical demonstrations utilising a real unmanned ground vehicle (UGV). The policy presented is capable of achieving multi-agent navigation tasks with a 95.83% accuracy in our testing environments, and has demonstrated emergent multi-agent behaviours. In more complex collaborative firefighting scenarios, the policy also demonstrated superior performance than baselines in reaching goals, e.g., navigating and extinguishing two fires with a 99.67% success rate, suggesting its strong potential for real-world deployment.

Details

Title: Multi-agent reinforcement curriculum learning for real unmanned ground vehicles
Authors: Timothy Mead - Cyborg Dynamics Engineering (Australia)
Zhe Wang - Griffith University
Ernest Foo - Griffith University
Jin Song Dong - National University of Singapore
Naipeng Dong - The University of Queensland
Ryan Ko - The University of Queensland
Abigail Koay - University of the Sunshine Coast, Queensland, School of Science, Technology and Engineering
Kien Nguyen - Queensland University of Technology
Yue Xu - Queensland University of Technology
Junae Kim - Defence Science and Technology Group, Brisbane, 4072, QLD, Australia
Stephen Bornstein - Cyborg Dynamics Engineering (Australia)
Zhe Hou (Corresponding Author) - Griffith University, Brisbane, 4111, QLD, Australia
Publication details: Engineering Applications of Artificial Intelligence, Vol.167, pp.1-13
Publisher: Elsevier Ltd
Date published: 2026
DOI: 10.1016/j.engappai.2026.113780
ISSN: 1873-6769; 0952-1976
Data Availability: Data will be made available on request.
Grant note: This research is funded by the Queensland Defence Science Alliance under the project QDSA_2023_OUT/063.
Organisation Unit: Healthy Ageing Research Cluster; School of Science, Technology and Engineering
Language: English
Record Identifier: 991206278502621
Output Type: Journal article

Metrics

1 Record Views

See more details