An Effective Hexapod Robot Control Design Based on a Fuzzy Neural Network and a Kalman Filter

Abstract

This study used a fuzzy neural network and a Kalman filter to control a hexapod robot. The robot could avoid obstacles while walking along a wall. The robot's posture was effectively adjusted, and high stability during movement was achieved. The angle between the robot and the wall was calculated by converting the distance values measured by ultrasonic sensors installed on the sides of the robot. For the fuzzy neural network, the input was the angular position and the output was the swing amplitudes of the robot's legs. The robot's forward movement direction was adjusted based on the difference between the swings of the legs on two sides, which allowed for obstacle avoidance in complex environments. The Kalman filter was used to obtain an accurate tilt angle in the robot by combining the advantages of a triaxial accelerometer and a gyroscope. The tilt angle was then further expressed via the movement direction of each leg. Each leg was adjusted using inverse kinematics, which allowed for the recovery of the robot's postural balance. Furthermore, this study proposes improved gait designs, which provided the robot with more effective responses to different landforms. The empirical results indicated that the proposed method increased the flexibility and mobility of the robot.