UniNeuro-HUDMOGA for advancing constrained optimization in double inverted pendulum

Abstract

Abstract This study aims for optimizing Double Inverted Pendulum (DIP) for swing then standing up within the stipulated time, and then can be stable even with any noise outside. The differences character outputs were applied to assess the robustness of proposed algorithm. This problem is worthy to examining, because in the controlling field, the stable position need in the matter of time; so then, the faster is the better sometime not necessary even dangerous. For example in comfortably of car’s suspension controlling system, the smoother suspension does not mean the faster stability, unless the comfortably never occurs. The difficulty is happened due to unknown model to optimize; it becomes the more complexity problem. In the real problem, even the model sometime cannot represent the whole of controlling system, especially in the controlling system. Facing the problem above, this study was succeeding for solving the problem via Uniform Design (UD) NeuralNetwork-Hybrid Multi Objectives Uniform Design Genetic Algorithm (UniNeuro-HUDMOGA). This proposed algorithm begins with 40 experiments designed by Uniform Design for generating predicting model (metamodel). Then, the metamodel is using for cost function in Hybrid UD Multi-objective Genetic Algorithm (HUDMOGA) within UD, Pareto filtering and Euclidean distance together applies for enhancing the searching GA performance for search the best setting. Finally, the input setting recommendation is confirming using DIP equipment. As the result, the DIP can be controlled as correlated with the specification set. These results prove that the proposed algorithm can be applied to control a complex system that requires multi objectives.