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

• Author / Creator

  Lemouchi, Rafik

• This work presents a novel approach to constructing site crane path planning using reinforcement learning and virtual-reality simulations. The approach involves a comprehensive simulation model that includes an agent, actions, states, environment, and a reward system. After extensive training over millions of episodes, the crane agent learns optimal path-planning techniques that improve lifting time, manage energy consumption, and enhance collision detection. The methodology consists of a multi-stage process that begins with creating a realistic virtual environment resembling a complex construction site. In this environment, the crane agent navigates through various scenarios, learning to adapt its path planning to dynamic changes, obstacles, and spatial constraints typical of construction projects. Dynamic changes are related to the placement of different equipment in the construction site and the movement of workers and materials within the site. Through continuous reinforcement learning, the agent refines decision-making, prioritizing efficient lift paths that minimize time and resource utilization, such as fuel consumption.
  The study results show significant improvements in terms of lifting time, lifting complexity, and energy consumption, which was achieved through reinforcement learning-based path planning for crane operations. The evolution of the crane agent from initial exploration to peak efficiency is demonstrated through cumulative rewards and decreasing simulation times. The study highlights the agent's ability to maneuver dynamically changing environments and optimize crane operations, showcasing the practicality and effectiveness of the proposed approach. Additionally, the integration of virtual-reality simulations enhances the agent training process by providing realistic scenarios and spatial awareness, which is crucial for crane operators in real-world settings. The study emphasizes the potential of combining advanced technologies such as reinforcement learning and virtual reality to revolutionize crane path planning, ultimately leading to safe, efficient, and environmentally sustainable construction practices.
  In conclusion, this research advances crane operations by introducing an innovative methodology that leverages state-of-the-art technologies to optimize path planning and enhance performance in dynamic construction environments.

• Subjects / Keywords

  • reinforcement learning
  • Mobile Crane
  • Path Planning
  • Virtual Reality

• Graduation date

  Fall 2024

• Type of Item

  Thesis

• Degree

  Master of Science

• DOI

  https://doi.org/10.7939/r3-r5nt-jc63

• License

  This thesis is made available by the University of Alberta Library with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.