Scribd
Upload
17 views1 page

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

Uploaded by

mohammad broomand
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
17 views1 page

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

Uploaded by

mohammad broomand
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 1

41st International Symposium on Automation and Robotics in Construction (ISARC 2024)

Reinforcement Learning-Enhanced Path Planning for


Mobile Cranes in Dynamic Construction Environments: A
Virtual Reality Simulation Approach
Rafik Lemouchi1, Mohamed Assaf1, Ahmed Bouferguene2 and Mohamed Al-hussein1

1
Department of Civil and Environmental Engineering, University of Alberta, Edmonton, Canada
2
Campus Saint-Jean, University of Alberta, Edmonton, Canada
[email protected], [email protected],[email protected], [email protected]

Abstract until later stages.


This paper introduces a novel approach to crane To address the path planning problem, many studies
path planning on construction sites through the have explored using automated planning tools and
utilization of Reinforcement Learning (RL) and information technology to enhance current practices. The
Virtual Reality (VR) simulations. The strategy initial works focused on using deterministic algorithms
includes a comprehensive simulation model that for path planning, as [3] outlined. To automate the path
incorporates an agent, actions, states, environment, planning task, they employed two heuristic search
and a reward system. After undergoing extensive methods, hill climbing and A*. Hill climbing involves an
training across millions of episodes, the crane agent iterative approach of adjusting a solution to minimize the
has acquired optimal path-planning techniques that distance or any desired cost function. In comparison, A*
enhance lifting time, manage energy consumption, uses a nodes approach, where it tries to find the shortest
and improve collision detection. The results highlight path between the start and end nodes using a cost function
the agent's impressive growth from initial exploration similar to hill climbing. However, these methods proved
to peak efficiency, represented by cumulative rewards time-consuming and often stuck in local optima instead
and evolving simulation times. The findings also of finding the optimal solution.
demonstrate the effectiveness of RL-based path Several studies have explored using metaheuristic
planning in maneuvering dynamic construction algorithms to improve crane path planning. [4] used an
environments and optimizing crane operations. ant colony to achieve collision-free path planning for
mobile cranes, while [5] employed genetic algorithms to
Keywords – plan lifts in complex environments. However,
Path Planning, Simulation, Reinforcement metaheuristic algorithms are only sometimes the optimal
Learning, Virtual Reality. solution and can be heavily influenced by initial
conditions. At the same time, other works, such as [6],
attempted to use hybrid approaches to tackle the issue,
1 Introduction with similar results as previous works.
The construction industry constantly evolves, aiming Recently, many researchers have turned to 3D/4D
to maximize efficiency and minimize costs. Over the past simulations and Building Information Modelling (BIM)
few years, off-site construction has gained traction due to to simulate and generate feasible solutions for lift
its ability to save time and money. This approach planning. [1] integrated 4D crane simulation and BIM to
involves transporting prefabricated modules to the manage operations on a construction site, while [7]
construction site for installation, making cranes an presented a data-driven crane management system for
essential component. industrial projects. Although BIM-based simulations
Despite numerous attempts to improve crane offer detailed visualization, they may struggle to handle
operations, outdated tools and planning methodologies dynamic scenarios, making them less adaptable. [8]
still need to be used. According to [1,2], current practices developed a methodology that enables automatic re-
involve lift engineers generating CAD-based 2D and 3D planning of lifting paths for robotized tower cranes in
simulations of various lift scenarios in a static and time- dynamic BIM environments to address this issue. They
consuming manner. The planning process is often trial used a GPU-based parallelization approach for discrete
and error, with better alternatives not being discovered and continuous collision detection. However, the

880

You might also like