Tuning a fuzzy inference system for nonlinear control applications using a hybrid metaheuristic algorithm

Fuzzy inference systems are well-suited to a wide range of control problems due to their ability to handle uncertainties in any nonlinear, complex system. The success of a fuzzy logic controller is determined by the parameters it uses, such as membership functions and rule base. To overcome this challenge, researchers have tried a variety of optimization strategies including metaheuristic algorithms. Because of its simple, resilient, and parallel searching properties, Particle Swarm Optimization (PSO) has found widespread use among metaheuristic algorithms. However, in high-dimensional space, this PSO technique may fall into a local optimum and has a poor convergence rate. This local optimum trapping issue is well addressed by the Simulated Annealing (SA) approach, which uses a random jump and metropolis acceptance in the solution space. This work describes a Hybrid Particle Swarm Optimization (HPSO) approach that incorporates the benefits of both PSO and SA synergistically. Simulated Annealing is used with PSO in the proposed HPSO to assist swarms in escaping local optima and driving them towards global optima. This HPSO is used to optimize the membership function parameters of a Sugeno-type fuzzy logic controller, which is tested on a benchmark cart and pole control problem, which mimics the self-balancing transporters. According to the findings of the tests, the HPSO-tuned fuzzy logic controller has a better control response than the classic PSO-tuned controller. In addition, in the HPSO-based technique, the cost function converged to a low value in every trial, but in the classic PSO-based approach, the solution converged to a local minimum in a few trials.