One of the main challenges in real-time robot control is its computational complexity and poor performance due to environmental uncertainties and nonlinearities. Interval Type-2 Generalized Fuzzy Hyperbolic Systems (IT2-GFHS) offer a simpler inferencing process but at the cost of increased number of inputs. However, numerous inputs could lead to decreased accuracy and lower efficiency. This article presents an Interval Type-2 Generalized Fuzzy Hyperbolic Neural Network System (IT2-GFHNS) that utilizes an Autoencoder neural network to reduce the number of inputs and improve the system performance. The effectiveness of the proposed method is first examined through simulations on a quadcopter and then experimentally implemented on an EMG-driven Robotic Follower, where the system identification is achieved using IT2-GFHNS. Additionally, the sliding mode is employed as a quadcopter control mechanism, and PID control is utilized for an EMG-driven robot. This method outperforms IT2-GFHS in terms of reducing tracking error, even with the increasing number of inputs.

Keywords