Investigation of Machine Learning for Clash Resolution Automation

9 pages•Published: June 9, 2021

Abstract

Various research work has recently investigated the utilization of Machine Learning for automating the process of clash resolution during design review and coordination of BIM models on construction projects. Literature review shows that current research work focuses on using Supervised Learning for automation of clash resolution. Individual implementation of Supervised Learning has its drawbacks. The automated model trained through Supervised Learning will only be able to resolve clashes similar in nature to those clashes used to train the model. This paper proposes a new approach that integrates Supervised and Reinforcement Learning to overcome these limitations. Reinforcement Learning will assist in overcoming the dependency of Supervised Learning on training data, while Supervised Learning will reduce the time for Reinforcement Learning by eliminating iteration with low rewards or illogical solution. The proposed approach will be able to assist industry practitioners in resolving clashes efficiently and effectively.

Keyphrases: Clash Resolution, Design Coordination, machine learning, Reinforcement Learning, supervised learning

In: Tom Leathem, Anthony J. Perrenoud and Wesley Collins (editors). ASC 2021. 57th Annual Associated Schools of Construction International Conference, vol 2, pages 228--236

Links:	https://easychair.org/publications/paper/n2t7
	https://doi.org/10.29007/n223