Test Program Generation Based on Multi-objective Evolutionary Algorithm

Existing test program evolution method uses single coverage metric to evaluate test programs in evolution process, and it can not guarantee the execution time index of the optimal solution. To cope with the problem, this paper proposes a test program generation approach based on multi-objective evolutionary algorithm (MOEA). It considers test program generation as a multi-objective optimization problem and both increasing coverage and reducing execution time are optimization goals. Through analyzing the feedback information from the simulator and using MOEA technology, it can automatically guide the direction of the new test generation. Experimental results on PKUnity UniCore32-2 microprocessor demonstrated that the optimal test program generated by the proposed method guaranteed the coverage requirements, was 12.92% of the traditional test program evolution method and 3.62% of hand-written test program set in execution time.