Towards the Automation of Last Planner System ®
Authors: Awe, M., Malhi, A., Mavengere, N., Budka, M. and Dave, B.
Journal: Proceedings of the International Symposium on Automation and Robotics in Construction
Pages: 1143-1152
eISSN: 2413-5844
DOI: 10.22260/ISARC2025/0148
Abstract:The increasing complexity and uncertainty in construction projects necessitate more efficient and adaptive construction management systems. This research studies the automation of the Last Planner System (LPS), a collaborative lean scheduling tool, by integrating predictive strategies to enhance scheduling accuracy, constraint resolution, and insights for decision-making. A conceptual framework is proposed for LPS automation while also representing the automation of LPS as an optimisation problem and determining the core aspects of LPS, such as communication management, constraint management, and planning and scheduling. This framework provides a foundation to which integration of predictive strategies in digitised Last Planner Systems is possible, focusing on process improvement, delay mitigation, and performance optimisation. A model that determines the constraint resolution duration was presented. This model used feature extraction techniques to determine constraint complexity score and preprocessed other features such as priority, location, and category to predict constraint resolution duration. The results indicated a Mean absolute error of 21.38 days (MAE) from the actual constraint resolution duration. Although the outcome and model can help in constraint management towards an automated LPS, the research identifies key limitations, including challenges in dataset quality, unfavourable sample sizes, and challenges with parsing complex textual and tabular data, highlighting areas for improvement. Future research should prioritise exploring emerging techniques, such as Table parsing (TAPAS) and Retrieval-augmented generation (RAG), to enhance the integration of data-driven strategies into Last Planner System workflows.
Source: Scopus