With the rapid development of intelligent construction technology, the dynamism and complexity of supply chain resource scheduling have significantly increased. Traditional scheduling methods are ill-suited to scenarios involving multi-entity collaboration, large fluctuations in resource demand, and numerous uncertainties in construction projects, leading to frequent problems such as resource waste and project delays. Current intelligent construction supply chain resource scheduling suffers from shortcomings such as lagging dynamic response, singular optimization objectives, and insufficient adaptation to multiple constraints, hindering efficiency improvements and cost control in construction projects. This paper first outlines the core elements and constraints of intelligent construction supply chain resource scheduling, constructing a dynamic scheduling problem model encompassing multiple dimensions including resource supply, demand fluctuations, and project schedule requirements. Second, it introduces reinforcement learning algorithms, designing a dynamic optimization framework based on state perception, action decision-making, and reward feedback, and achieving autonomous learning and iterative optimization of scheduling strategies through deep reinforcement learning networks. Finally, a multi-scenario simulation experimental platform is built to verify the scheduling performance of the model under different project scales and uncertainties. Experimental results: Experiments show that the proposed RL model has a lower overall scheduling cost than GA, PSO, and Rule-Based, improves resource utilization to 89.3%, and has the shortest task completion time with dynamic response latency controlled within 0.8s, which is significantly better than the benchmark model.
Liu et al. (Thu,) studied this question.