A REINFORCEMENT LEARNING AND THE NORTHERN GOSHAWK OPTIMIZATION ALGO- RITHM FOR FLEXIBLE JOB SHOP SCHEDULING PROBLEM

This paper introduces northern goshawk optimization, a novel global search strategy for the flexible job shop scheduling problem. It uses two-stage encoding and random-key-based encoding to transform individual position vectors into flexible job shop scheduling problem solutions. To improve local search, reinforcement learning is integrated, converting the flexible job shop scheduling problem into a Markov decision process with 10 states and 6 rules. A reward function based on optimal completion time guides the search. The proposed hybrid northern goshawk optimization-Q-learning-state-action-reward- state-action framework combines global and local search strengths. Experiments on standard datasets show the algorithm's superior performance, validating its effectiveness and practicality in solving the flexible job shop scheduling problem and real-world production scheduling problems.