NMPC Based-Trajectory Tracking and Obstacle Avoidance for Mobile Robots

This paper presents the design of a Nonlinear Model Predictive Controller (NMPC) for a wheeled Omnidirectional Mobile Robot (OMR) in order to track a desired trajectory in the presence of previously unknown static and dynamic obstacles in the environment around the robot. A laser rangefinder sensor is used to detect the obstacles where each obstacle occupies numerous points of every sensor reading. The points that belong to each obstacle are then clustered together using the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm. This research introduces a novel approach to represent obstacles as multiple rotated ellipses, enabling a more accurate representation of complex obstacle shapes without overestimating their boundaries, thereby allowing the robot to navigate through narrow passages. CoppeliaSim robotic simulator is utilized to create the virtual simulation environment as well as simulate the OMR dynamics. MATLAB with the help of the CasADi toolbox is used for the process of the laser rangefinder readings and the implementation of NMPC, respectively. To validate the effectiveness and robustness of the proposed approach, three simulation scenarios are conducted, each involving distinct trajectories and varying densities of static and/or dynamic obstacles. The proposed control architecture exhibits remarkable performance, enabling the OMR to effectively navigate through narrow passages and avoid multiple static and dynamic obstacles while closely adhering to the desired trajectory.