An integrated game-theoretic and reinforcement learning modeling for multi-stage construction and infrastructure bidding

Construction and infrastructure bidding is a highly competitive and complicated process that entails various uncertainties faced by contractors. The situation is more complex in multi-stage bidding (MSG) where general contractors must deal with the complexity of accounting for the bids of their subcontractors and face a greater threat of falling prey to the winner’s curse (i.e. situation where the winning contractor underestimates the actual cost of the project). Existing research efforts have tackled the issue of the winner’s curse in MSG from the general contractor’s perspective. However, there is a lack of research in developing bidding models that simultaneously aid both general contractors and subcontractors in determining their bid value to mitigate the winner’s curse in MSG. This paper fills this knowledge gap. The authors utilized an interdependent game theory (GT) and reinforcement learning (RL) approach, that includes: formulation of MSG framework; incorporation of two RL algorithms, namely the multiplicative weights and the modified Roth-Erev, to be utilized by subcontractors in preparation of their bids; utilization of MSG game-theoretic bid function for the preparation of the general contractors' bids for the whole project; development of the MSG-GTRL model; and testing the MSG-GTRL model through simulating various bidding scenarios using a combination of actual and synthetic dataset of infrastructure projects. Results show that integrating GT and RL in MSG bidding enables general contractors and their subcontractors to simultaneously improve their financial state by minimizing the occurrence of negative earnings, and thus, avoiding the winner’s curse in their respective portions of projects.