Improved Model Predictive Control System Design and Implementation for Unmanned Ground Vehicles

Autonomous ground robots autonomously are being used inthe places where it is very hazardous for human beings to reach and operate,such as nuclear power plants chemical industries. The aim of the researchpresented here develop a control system that enables such robotsnavigate with various sensors depth camera, 2D scanninglaser, 3D Lidar, GPS, IMU. controller uses current positionmeasured using on Husky A200, given waypoints thedestination. Then calculates best possible route based recentevents provided IMU data GPS. Model Predictive Control (MPC)improves robot’s motion, by path planner trajectorygeneration. use global reference frame planned createthe appropriate actions required follow motion planner’sdirection. depends active sensor locationsand size obstacles. Then, feasible generated sensordata. desired trajectory consists set fit in 3rd-orderpolynomial. They determine path’s feasibility robot’sdynamics series points certain velocity andacceleration profile. MPC adjusts lateral, longitudinal, yawmotions approximates continuous discrete paths commandbehaviors. kinematic model robot, dynamic modelfor transient steady-state characteristics. camera captures imagesand other types processed through computational framework tobuild machine learning models. TensorFlow deep toidentify classify objects around Husky. This research haslimitations linear LQR method. Also onvehicle models, vehicle considered this considers aconstant value describe slope most region. Detaileddiscussion development major design factor has beenemphasized logical steps MPC.