Experimental Fuzzy Logic Controller Type 2 for a Quadrotor Optimized by ANFIS

Abstract Quadrotors are nonlinear systems that can be controlled by human experts. Since the mathematical models of Quadrotors are quite complex, the expert knowledge may be one way to come to a solution for controlling Quadrotors. However, human experts make occasionally mistakes, and thus some linguistic rules used in the controller may be false or redundant. Hence, fuzzy logic type 2 optimized by ANFIS (Adaptive Neuro-Fuzzy Inference Systems), which can deal with uncertainties, could be applied to control Quadrotors. ANFIS can optimize the number of linguistic rules and the domain of membership functions could be adjusted automatically. In addition, ANFIS should also be capable of identifying bad rules so the digital system (micro-controller) can decrease the computational resources required for implementing the fuzzy logic controller type 2. The controller designed can accomplish excellent experimental results when it is reduced by an ANFIS system. To confirm robustness in the fuzzy logic controller a noise signal was added in the position control loop for the Quadrotor. Besides, a comparison between fuzzy logic controller type 2 tuned by an expert and Fuzzy Logic type 2 optimized by an ANFIS is illustrated. Experimental results confirmed the good response reached when fuzzy logic type 2 optimized by ANFIS is deployed.