Publications / 2024 Proceedings of the 41st ISARC, Lille, France

Reinforcement Learning-Enhanced Path Planning for Mobile Cranes in Dynamic Construction Environments: A Virtual Reality Simulation Approach

Rafik Lemouchi, Mohamed Assaf, Ahmed Bouferguene, Mohamed Al-Hussein
Pages 880-887 (2024 Proceedings of the 41st ISARC, Lille, France, ISBN 978-0-6458322-1-1, ISSN 2413-5844)
Abstract:

This paper introduces a novel approach to crane path planning on construction sites through the utilization of Reinforcement Learning (RL) and Virtual Reality (VR) simulations. The strategy includes a comprehensive simulation model that incorporates an agent, actions, states, environment, and a reward system. After undergoing extensive training across millions of episodes, the crane agent has acquired optimal path-planning techniques that enhance lifting time, manage energy consumption, and improve collision detection. The results highlight the agent's impressive growth from initial exploration to peak efficiency, represented by cumulative rewards and evolving simulation times. The findings also demonstrate the effectiveness of RL-based path planning in maneuvering dynamic construction environments and optimizing crane operations. Future work involves conducting a detailed comparison with real-life scenarios to bridge simulation and reality, ensuring the applicability of the model in practical construction settings.

Keywords: Path Planning, Simulation, Reinforcement Learning, Virtual Reality.