Crane Mat Layout Optimization Based on Agent-Based Greedy and Reinforcement-Learning Approach

The growing prevalence of modular construction, while it offers benefits in terms of productivity and sustainability, has led to increased use of heavy mobile cranes and related resources on construction sites. One significant resource associated with crane use is the crane mat, which offers practical mobile crane ground support against poor soil-bearing capacity. Due to increased use of crane mats, crane mat layout plans/drawings have become increasingly significant in today’s construction industry. The present work describes an automated crane mat optimization framework for preparing crane mat layout plans/drawings built on an agent-based greedy algorithm and reinforcement learning. The proposed framework employs these approaches to achieve the maximum area with the minimum number of crane mats. The proposed framework is found to decrease the time required for preparing a crane mat layout plan/drawing (approximately 97% time saving) with more uniform and efficient crane mat planning outcomes (approximately 63% crane mat material reduction).

The research presented in this manuscript examined the largely unexplored topics of crane mat layout optimization and layout preparation, proposing an agent-based greedy algorithm and reinforcement learning (RL) approach for automated crane mat layout optimization as an innovative approach to developing sustainable crane mat layouts. This approach takes into account the site constraints and can be applied to mitigate crane mat crowding on construction sites. Crane mat optimization is applied using both methods (greedy and RL) to achieve the maximum area covered with the minimum number of crane mats used. The results demonstrate that the practitioner time spent preparing a crane mat layout plan/drawing can be reduced considerably, with more uniform and cost-effective crane mat optimization outcomes. Another outcome of this research is that developed approaches reduce the crane mat requirements compared with the outputs generated by the manual approach, thereby reducing the ${\mathrm{CO}}_2$ emissions associated with crane mat manufacturing and usage.