Memetic Compact Differential Evolution for Cartesian Robot Control

摘要

This article deals with optimization problems to be solved in the absence of a full power computer device. The goal is to solve a complex optimization problem by using a control card related to portable devices, e.g. for the control of commercial robots. In order to handle this class of optimization problems, a novel Memetic Computing approach is presented. The proposed algorithm employs a Differential Evolution framework which instead of processing an actual population of candidate solutions, makes use of a statistical representation of the population which evolves over time. In addition, the framework uses a stochastic local search algorithm which attempts to enhance the performance of the elite. In this way, the memetic logic of performing the optimization by observing the decision space from complementary perspectives can be integrated within computational devices characterized by a limited memory. The proposed algorithm, namely Memetic compact Differential Evolution (McDE), has been tested and compared with other algorithms belonging to the same category for a real-world industrial application, i.e. the control system design of a cartesian robot for variable mass movements. For this real-world application, the proposed McDE displays high performance and has proven to considerably outperform other compact algorithms representing the current state-of-the-art in this sub-field of computational intelligence.