Abstract
Multi-link inverted pendulum systems pose intricate challenges in control theory and robotics, requiring precise dynamic parameter identification to achieve stability and robustness in control strategy design. We present a novel and efficient experimental identification procedure formulated as an optimization problem based on simple short-term datasets and metaheuristic global optimizers. We use a training dataset for identification and validation dataset to evaluate and analyze the obtained results. The study incorporates three distinct global optimization techniques, namely Stochastic Fractal Search (SFS), Growth Optimizer, and Differential Evolution (DEoptim), selected as candidates to handle the identification of multi-link pendulums and similar extremely demanding optimization jobs to be used when controlling modern mechatronic systems. We illustrate that DEoptim dominates over other global optimizers in several aspects. The proposed identification procedure is innovative, adaptable, and simple, relying solely on selected measurable signals sans further signal processing. Its versatility makes it a valuable tool for parameter identification in diverse domains. The results are supported by experiments with the laboratory triple pendulum setup and simulation experiments on a virtual quadruple inverted pendulum.