Neural Controller Design of Unmanned Ground Vehicle with Four-Wheel Independent Drive in Agricultural Farming

The application of unmanned ground vehicles (UGVs) in smart agriculture is becoming increasingly prominent, UGV is often used to carry the tools for planting and taking care of crops. Due to the characteristics of farming in agriculture, UGVs are usually four-wheel independent drive autonomous ground vehicles that has eight degrees of freedom (dof). Hence, precise trajectory tracking control for UGVs is a difficult issue for the hardware architecture of the controller because it has to handle a large number of calculations collected from environment. So, it is essential to design a controller which has a not too complicated configuration and small computation volume but still ensures good real-time performance without requiring high hardware configuration to enhance the cultivation trajectory tracking accuracy for UGV. This article presents a simple neural network controller based on the PID reference controller to track the cultivation trajectory for UGV. To tracking control for UGV, the three-layer backpropagation neural network controller is conducted. The neural network is trained with a sample input - output data set such that the control signal error should be small as possible. To illustrate this approach, the tracking error model of the eight dof UGV was established, on that basis, a tracking simulation for the NURBS cultivation trajectory was carried out. The feasibility of the obtained controller is investigated through the tracking error of the UGV's cultivation trajectory, which is lower than 0.0782 m, and the linear and angular velocity errors are lower than 0.0017 m/s and 0.0093 rad/s, respectively.