Self-Balancing Vehicle Based on Adaptive Neuro-Fuzzy Inference System

Abstract

The scope of this research is to design and fuse the sensors used in the self-balancing vehicle through Adaptive Neuro-Fuzzy Inference systems (ANFIS) algorithm to optimize the output. The self-balancing vehicle is a wheeled inverted pendulum, which is extremely complex, nonlinear and unstable. Homogeneous and Heterogeneous sensors are involved in this sensor fusion research to identify the best feasible value among them. The data fusion algorithm present inside the controller of the self-balancing vehicle makes the inputs of the homogeneous sensors and heterogeneous sensors separately for ameliorate surrounding perception. Simulation is performed by modeling the sensors in Simulink. The outcomes specifies that the data fusion algorithm allocates minimal root mean square error (RMSE) and mean absolute percentage error (MAPE) when analyzed and compared with that of every sensor in the system. Finally, the output signals of these sensors are examined and viewed along with noise signal and the actual signal is isolated from the noise signal by applying extended Kalman filter. This propounded technique of ANFIS based fusion algorithm has improved RMSE for both homogeneous sensors and heterogeneous type sensors. Robotic systems may execute several control strategies in various proximity levels based on the performance of the data fusing algorithm.