Path Following Control Of Wheeled Mobile Robots Based On Model Predictive Control

Path following is an important kind of motion control problem as same as the setpoint stabilization and trajectory tracking in the control theory. Compared with trajectory tracking, path following controller needs to determine the information of the reference path as well as the system dynamics. Path following is widely present in computer numerical control(CNC) machine tools, motion control of mobile robot, course control of airplane and ship, intelligent autonomous cars and other fields. Therefore, studying path following problem not only can enrich the theoretical results of motion control, but also can meet the requirements of the motion control technique, which has important theoretical and engineering significance.Wheeled mobile robot is a typical example of the nonlinear constrained system, which is chosen as the motion control plant in this thesis. Compared with other control schemes,nonlinear model predictive control(NMPC) not only has feedback mechanism, but also can handle the system state/input constraints. So NMPC is applied to path following control for wheeled mobile robots. This thesis is organized as follows to accomplish the design of following control system for wheeled mobile robots: problem description, establishment of mathematical model of the controlled plant, controller design according to the control goals, the stability analysis of the closed-loop system and simulation verification.Firstly, the description of path following problem is given and the kinematic model of unicycle mobile robot is established, then the control task is proposed. The unicycle mobile robot is composed of two independent drive rear-wheels and a universal wheel,in which the simple mechanistic structure makes the movement flexible and reliable. In principle, the unicycle mobile robot is a multiple-input and multiple-output nonlinear system with state and input constraints.To deal with nonlinear dynamics and constraints, this thesis proposes a model predictive control scheme, in which a nonlinear optimization problem solved online is involved.Both the recursive feasibility of the optimization problem and the system performance are discussed. This thesis computes the terminal ingredients that guarantee the feasibility of proposed NMPC based on the nonlinear polyhedral description and linear matrix inequalities, and gives the comparison and analysis between the NMPC scheme with terminal inequality constraint and the NMPC scheme with terminal equality constraint.The simulation results show that both the NMPC scheme with terminal inequality constraint and the NMPC scheme with terminal equality constraint can make the robot follow the reference path under the admissible control inputs. But the NMPC scheme with terminal inequality constraint can only solve the problem with smooth paths, and the online computation burden of the NMPC scheme with terminal equality constraint is much heavier. In accordance with the fact that reference path is time varying, a MPC scheme with time varying terminal sets is proposed which can deal with non-smooth path and have reasonable computational time.