Look-ahead planning is the stage in construction planning where information from diverse sources is integrated and plans developed for the next six/eight weeks. Poor planning of construction site activities at this stage often results in cost overruns and schedule delays. This work presents a novel Look-Ahead Schedule (LAS) generation method, which uses reinforcement learning and linked-data based constraint checking within the reward, to address the issues associated with manual look-ahead planning and help construction professionals efficiently plan construction activities at this stage. Our proposal can generate conflict-free LAS significantly faster than conventional methods, demonstrating its capability as a decision support tool during look-ahead planning meetings. Therefore, this paper extends existing knowledge in the construction informatics domain by demonstrating the application of reinforcement learning to aid data-driven look-ahead planning.

In the construction sector, complex constraints are not usually modeled in conventional scheduling and 4D building information modeling software, as they are highly dynamic and span multiple domains. The lack of embedded constraint relationships in such software means that, as Automated Data Collection (ADC) technologies become used, it cannot automatically deduce the effect of deviations to schedule. This paper presents a novel method, using semantic web technologies, to model and validate complex scheduling constraints. It presents a Linked-Data based Constraint-Checking (LDCC) approach, using the Shapes Constraint Language (SHACL). A prototype web application is developed using this approach and evaluated using an OpenBIM dataset. Results demonstrate the potential of LDCC to check for constraint violation in distributed construction data. This novel method (LDCC) and its first prototype is a contribution that can be extended in future research in linked-data, BIM based rule-checking, lean construction and ADC.