Reliability allocation optimization problem of a complex mechatronic system is a highly nonlinear constrained optimization problem, and hence solution to this kind of problem is of NP-hardness even with moderate scale. Due to the nonlinearity combined with multiple local extreme values, traditional optimization techniques fail to arrive at the global or nearly global optimal solution to the problem. Genetic algorithm incorporated with neighboring domain traversal searching technique is utilized in this paper to solve the complex mechatronic system reliability optimization allocation problem. Reliability allocation optimization of the life-support system in a space capsule, being a typical non serial-parallel system, is specifically demonstrated to show the satisfactory convergence performance as well as the important practical value of hybrid genetic algorithm. The simulation results show that the proposed method may gain better precision in solving the complex mechatronic system reliability optimization problem.