Application of Genetic Algorithm with Biased Random Keys for Dynamic Job Shop Problems

The dynamic job shop scheduling problem (DJSSP) captures realistic production environments with machine breakdowns, new job arrivals, and processing-time changes. This paper presents a reactive rescheduling approach based on the Biased Random-Key Genetic Algorithm (BRKGA), preserving the separation between problem-independent evolutionary logic and a problem-specific decoder. The method adapts BRKGA to the dynamic setting through a relativepriority decoder, hybrid initialization of the initial population, and a warm-start mechanism that partially reuses the population across events while keeping completed operations fixed and reoptimizing only the residual schedule. The experimental protocol includes static and dynamic instances, with automatic parameter tuning through irace for both BRKGA and IHKA, and 30 runs per instance for each algorithm.