Optimal Design of Infrared Motion Sensing System Using Divide-and-Conquer based Genetic Algorithm

This paper studies the optimal design of an infrared motion sensing system for human motion localization in the context of human-following robots. Specifically, we aim to find the optimal number and placement of bearing-sensitive pyroelectric infrared (PIR) sensor arrays for improving localization performance. This optimal design leads to a multiobjective, mixed-integer-discrete-continuous and variable-dimensional optimization problem, which prevents from using conventional multiobjective optimization techniques including genetic algorithm (GA). This paper explores the use of divide-and-conquer based GA in solving this optimal design problem. The proposed approach consists of three steps: firstly, following divide-and-conquer principle, the optimal design problem is decomposed into a set of sub-optimization problems; then the sub-optimization problems are solved with standard GA; finally, the optimal solution is found through fusing the resulting solutions of sub-optimization problems. The proposed design approach is illustrated with a design example, and verified with experimental studies.