Machine Learning for Construction Scheduling
Exploring innovative opportunities for flexible project schedule design and execution.
After the launch of the research partnership between the University of Cambridge’s Construction Information Technology (CIT) Lab, nPlan and Kier, the Cambridge team continues to make progress in the exploration of the integrated schedule execution framework comprised of process benchmarking and delay-prediction tools. This research programme, sponsored by the InnovateUK grant, is titled “AI-Optimised Pathways for Schedule Execution (AI-OPSE)”. The project team aims to improve project management by developing an innovative schedule-learning platform which applies data science and machine learning to thousands of historic project schedules. The project will offer a unique and scalable solution for improving reliability and confidence in project planning. There are seven phases in this 24-month industrial research programme. The Cambridge team is in charge of framework development on learning optimal sequences of tasks and learning comparative project performance.
At present, the Cambridge team is designing and developing the prototype framework that uses an identified artificial intelligence (AI) model to read sequences of tasks based on high-level human input. The expected system output will be a probabilistic estimation of delays which can help to train a second AI model to rearrange the existing tasks for the purpose of minimising risks and possible delays. This novel approach will empower the automation of benchmarking and determination of task efficiency and project performances.
Tags: Planning , Analyse