Non-holonomic agricultural robot with neural network on-line learning controller

The present study represents agricultural applications of non-holonomic mobile robots. Agricultural robots are significantly affected by various disturbances such as a loading capacity and modeling errors of robots. Abundant studies of mobile robots using online learning have investigated real-time elimination of control errors that are associated to inaccurate modeling of robots and controls of disturbance effects by online learning using neural network. However, a certain problem from errors of online learning may occur in case of the robot that is located on out of tracking path. The above problem may be resulted in the difference by control errors occurring between desired values and current values. Therefore, stepwise optimization of robot control with desired values should be necessary. Online learning for agricultural mobile robots is possibly performed with accurate calculation of control errors using the stepwise optimization of desired values as a standard. In summary, the present study demonstrates a reference robot is used to calculate accurate control errors for non-holomic mobile robot that is driven by pulse. The control error of the non-holonomic mobile robot through online feedback-error learning is almost 1.3% in agricultural application.