Abstract—Autonomous self-driving for car-like mobile robots are taken into account for various applications such as spraying agrochemical, mowing grass, military operations, commercial delivery service, and so on. In this paper, a simple kinematic bicycle model of the mobile robot is introduced and Model Predictive Control (MPC) mechanism is also described. In addition, this paper derives the hybrid control architecture which is based on Deep Neural Network (DNN) based learning and Model Predictive Control (MPC) scheme for the self-driving of the car like mobile robot. Computational Simulation results show that the proposed controller architecture can successfully generate its output such as the steering angle, velocity, and acceleration in terms of the robot kinematics and produce a desired trajectory during self-driving. Finally, the effectiveness of DNN MPC model is presented by comparing with Dynamic Window Approach (DWA).

Index Terms—autonomous self-driving, deep neural network, model predictive control, artificial neural network, kinematic bicycle model, car-like mobile robot

Cite: Kiwon Yeom, "Design of Deep Neural Network Based Model Predictive Controller for a Car-like Mobile Robot," International Journal of Mechanical Engineering and Robotics Research, Vol. 11, No. 8, pp. 606-613, August 2022. DOI: 10.18178/ijmerr.11.8.606-613