Multi-Mode Resource-Constrained Project Scheduling Based on a Combined Nsga Ii and Q-Learning Algorithm

Project scheduling is one of the core elements of project management, which is essential for engineering-to-order manufacturing companies to improve productivity, reduce costs and shorten project finish time. This paper explores the multi-mode project scheduling problem under resource and finish time constraints and establishes a mathematical model with the objectives of shortening the project cycle and improving the load balance of resources. Considering that the start time selection of each job is independent of each other and is consistent with the characteristics of the Markov decision process, a two-layer iterative algorithm is proposed to solve the model based on NSGA Ⅱ and Q-learning algorithm. NSGA Ⅱ algorithm generates a combination of modes, and its fitness function searches for job time selection in each mode using the Q-learning algorithm. The proposed algorithm’s performance priority is verified by comparing it with the classical NSGA Ⅱ, Particle swarm optimization, and Ant colony optimization algorithms