Multilevel Redundancy Allocation for SEU Mitigation of DSP Codes Based on OMA

Abstract

Single event upset (SEU) is one of the major factors affecting the reliability of digital signal processor (DSP) cored software system in aerospace equipment. The current methods of DSP mitigation involve ${N}$ -modular redundancy refresh, Hamming code, and so on. Based on the idea of modular redundancy, this paper turns the SEU mitigation of DSP codes into an optimization problem to find out the multi-level redundant allocation solution considering the structure division, cost evaluation, and reliability modeling of DSP codes. The optimal mitigation solution is determined by optimized memetics algorithm (OMA) with modified crossover operator and local search operator in order to improve the optimization speed and convergence stability. At the same time, by distributing the redundancy of every module according to the optimal solution of the OMA, the DSP projects can get a higher degree of reliability under cost constraints.