Trajectory Tracking of a Spherical Robot Based on an RBF Neural Network

Abstract

This paper deals with trajectory tracking problem of a spherical mobile robot, BHQ-1. First, a desired velocity is obtained by proposing a PD controller based on the kinematics. Then a PD controller with an RBF (Radial Basis Function) neural network is proposed based on the desired velocity and the inexact dynamics. The weights of the RBF network are designed with an adaptive rule based on the tracking error, and hence the network can compensate the uncertainties of the dynamics more effectively. Stability is presented via Lyapunov Theory and simulation results are provided to illustrate the tracking performance.

Dynamic model based adaptive sliding mode trajectory tracking control of pendulum-driven spherical mobile robot

“…Otani et al [18] designed a Back-Stepping controller based on dynamic model, which could track expected trajectory in the presence of certain error. Zheng et al [15] proposed a PD controller with a RBF neural network of spherical robot considering dynamics of modelling error, and verified its good tracking performance. Roozegar et al [21] designed a heuristic fuzzy control law and a PID control law for spherical robot respectively, and verified their effectiveness by controlling a spherical robot to climb a slope.…”

“…Some researchers [12][13][14] designed controllers based on decoupled dynamic models neglecting the dynamic interaction between forward rolling motion and turning motion of spherical robot, but these controllers can only use the two motions alternately to approximately meet the assumptions in the process of tracking target trajectory, because spherical robot is an under-actuated and strong coupling system. Some researchers [15][16][17][18][19] chose the dynamic model of the homogeneous spherical shell as the research object ignoring the internal structure of spherical robot. However, the internal driving unit is the core component of spherical robot for motion and rapid recovery to stability, if ignoring it, the motion characteristics of spherical robot will be changed.…”

Design and validation of a novel adaptive motion control for a pendulum spherical robot

“…Zadeh F. K. et al [24] designed an efficient LQR to eliminate the oscillation in a SR's motion and improve the system's stability. Zhan Q. et al [25][26][27][28][29] utilized backstepping, the neurodynamic shunt model, the cerebellar model joint control model, and RBF neural network to improve the controller's disturbance immunity.…”