Multi-Project Optimization with Multi-Functional Resources by a Genetic Scheduling Algorithm

In this paper we show how a genetic scheduler algorithm can be applied to solve a hard multi-project optimization problem with shared resources. The resources work in multiple operation modes, so they can substitute each other (but with different efficiency). We consider processes which have quite complex structure, i.e., it allows the existence of parallel sub-processes. This problem is extremely complex, there is no chance to get the optimal solution in reasonable time. The proposed algorithm intends to find a near-optimal solution, where the goal of the optimization is the minimization of the makespan of the schedule. We present the genetic operations of the algorithm in detail. We fill the pool of populations only with feasible solutions, but making possible the discovery of the whole search space. The feasibility of a schedule is ensured by excluding time-loops regarding the sequence of the tasks both in their process and in the queue of their resource. We executed several tests for determining the (hopefully) optimal parameters of the algorithm regarding the number of generations, the population size, the crossover rate and rate of the mutation. We applied the algorithm for many problem classes where the parameters of the input are fixed or randomly chosen from some interval.