The advent of Building Information Modelling (BIM) has fundamentally altered the way how building projects are approached. Nevertheless, it is still the case that conflicts during the collaboration process are an unavoidable consequence, resulting in inefficiencies. The widespread adoption of BIM model checker has significantly contributed to the detection of conflicts. However, the automation of conflict resolution remains a nascent endeavor, and the reliance on manual communication among designers can be a significant time investment. To address this shortcoming, this research presents a framework for training a Reinforcement Learning (RL) agent, employing the Proximal Policy Optimization (PPO) algorithm, in the integrated real BIM environment to automate the conflict resolution process, with a particular focus on geometric conflicts. Three experiments, each focusing on a different type of conflict, were conducted to investigate the feasibility of the proposed framework. The results were analyzed, and the limitations were discussed.     «

The advent of Building Information Modelling (BIM) has fundamentally altered the way how building projects are approached. Nevertheless, it is still the case that conflicts during the collaboration process are an unavoidable consequence, resulting in inefficiencies. The widespread adoption of BIM model checker has significantly contributed to the detection of conflicts. However, the automation of conflict resolution remains a nascent endeavor, and the reliance on manual communication among desig...     »