Amphibious robot's trajectory tracking with DNN-Based nonlinear model predictive control

Amphibious robots are being deployed in field environments where they are required to handle environmental disturbances and systemic uncertainties. Efficient and accurate control strategies can guarantee high performance of the robot's trajectory tracking tasks. In this paper, we first contribute a well design deep neural network (DNN) as a precise black-box kinematic model of the amphibious robot. Then, we design a DNN based nonlinear model predictive controller (DNN-NMPC) which obtains the robot's real-time moving command by iterative optimization. To verify the proposed method's performance in amphibious robots' trajectory tracking tasks, a Gazebo based simulation platform has been built and several comparative simulations have been carried out. The simulation results indicate the proposed controller is superior to the basic controller in the robot's tracking efficiency and accuracy.