Implementation of AI for construction scheduling

A construction schedule is one of the most essential documents to manage and measure a project’s performance. Its completeness and correctness have a direct impact on the project’s success. However, the traditional methodology to create the schedules presents several flaws; such that statistics show that more than 75% of the projects worldwide do not finish in the duration predicted by their original plan.
This dissertation researched the mainstream scheduling practice, through literature review and a case study based on Germany and found that it extensively relies on the human scheduler capabilities to generate and process all the data necessary for a schedule. Two main issues were identified:
- Firstly, there is not a productivity rate (PR) database that supports the calculation of activities duration. And, this leads to base the estimates on personal experience.
- Second, the constraints to which a project is subject are not clearly indicated in the schedule and they are all resolved based on precedence relations.
Focusing on the first issue, a literature review of factors affecting the productivity was conducted and these influences were categorized in objective and subjective factors. This classification was made with the aim of identifying which factors could be controlled at the time of measuring productivity. Then, a methodology to measure consistent productivity data at the activity level was developed and implemented in a web-based application to facilitate the on-site data collection. This methodology was used to collect actual PRs from the case study and to start building a database that can be used for future estimations.As for the second issue, a new scheduling platform based on the tri-constraint method (TCM) was used to try its potential to better resolve and communicate the project’s constraints. This platform is a parametric based scheduling software, that by using the TCM solves the project’s limitations in terms of physical, spatial and resources constraints. This new technology was used to create various schedules for the case study, and they were compared to the original (Gantt Chart) schedule. It was found that by using the TCM the communicability of the schedule was improved due to its 5D capabilities. Furthermore, the schedules created by using TCM and the productivity data collected on-site resulted in longer durations than the original estimation, but they seem to approach more closely to the real progress observed on the construction site.
Additionally, this dissertation provides a review of the strengths and debilities of the TCM software. As well as some recommendations to prepare projects and firms to benefit from the new scheduling technologies and move towards a more standardized planning practice.