Non-linear model predictive control with single-shooting method for autonomous personal mobility vehicle

The advancement of autonomous vehicle technology has markedly evolved during the last decades. Reliable vehicle control is one of the essential technologies in this domain. This study aims to develop a proposed method for controlling an autonomous personal mobility vehicle called SEATER (Single-passenger Electric Autonomous Transporter), using Non-linear Model Predictive Control (NMPC). We propose a single-shooting technique to solve the optimal control problem (OCP) via non-linear programming (NLP). The NMPC is applied to a non-holonomic vehicle with a differential drive setup. The vehicle utilizes odometry data as feedback to help guide it to its target position while complying with constraints, such as vehicle constraints and avoiding obstacles. To evaluate the method's performance, we have developed the SEATER model and testing environment in the Gazebo Simulation and implemented the NMPC via the Robot Operating System (ROS) framework. Several simulations have been done in both obstacle-free and obstacle-filled areas. Based on the simulation results, the NMPC approach effectively directed the vehicle to the desired pose while satisfying the set constraints. In addition, the results from this study have also pointed out the reliability and real-time performance of NMPC with a single-shooting method for controlling SEATER in the various tested scenarios.