Modified Differential Evolution in the Load Balancing Problem for the iFDAQ of the COMPASS Experiment at CERN

In general, state-of-the-art data acquisition systems in high energy physics experiments must satisfy high requirements in terms of reliability, efficiency and data rate capability. The paper introduces the Load Balancing (LB) problem of the intelligent, FPGA-based Data Acquisition System (iFDAQ) of the COMPASS experiment at CERN and proposes a solution based on genetic algorithms. Since the LB problem is N P-complete, it challenges analytical and heuristic methods in finding optimal solutions in reasonable time. Differential Evolution (DE) is a type of evolutionary algorithms, which has been used in many optimization problems due to its simplicity and efficiency. Therefore, the Modified Differential Evolution (MDE) is inspired by DE and is presented in more detail. The MDE algorithm has newly-designed crossover and mutation operator and its selection mechanism is inspired by Simulated Annealing (SA). Moreover, the proposal uses an adaptive scaling factor and recombination rate affecting the exploration and exploitation of the MDE algorithm. Thus, the MDE represents a new efficient stochastic search technique for the LB problem. The proposed MDE algorithm is examined on two LB test cases and compared with other LB solution methods.