Mingi Ji, Earthwork Planning in Various Environment via Reinforcement Learning, Master's Thesis, Department of Industrial and Systems Engineering, KAIST, 2018

Abstract

Earthwork is the leading work of most construction projects and one of the most important tasks in construction process management. For the construction process management, there are some different approaches such as optimizing construction with either mathematical methodologies or heuristics with simulations. This paper proposed a simulated earthwork scenario and an optimal path planning for the simulation using a reinforcement learning. For reinforcement learning, excavator agent its behavior policy using Q-learning and deep reinforcement learning. The simulation result shows that learning with Q-learning and deep reinforcement learning can reach the optimal planning for a training simulated earthwork scenario. In addition, an earthwork planning path was created by applying the trained excavator agent to simulated environment that was not used for training. Excavator agent trained from deep reinforcement learning can reach near optimal planning. This planning could be a basis for an automatic construction management. 

@masterthesis{Ji:2018,

author = {Mingi Ji},

advisor ={Il-Chul Moon},

title = {Earthwork Planning in Various Environment via Reinforcement Learning},

school = {KAIST},

year = {2018}

}