الفهرس 
Development of Fuzzy Logic Based Model Predictive Control for Path Tracking of the Autonomous Vehicle
SummaryOnly 14 pages are availabe for public view
 |  | from | 131 |  |  |
| | | from | 131 | | |
Abstract
This thesis introduces apath tracking control strategy for autonomous vehicles. The proposed strategy applies a fuzzy-based switching system with model predictive control on a full nonlinear vehicle model. This model includes a four-wheel vehicle dynamic and a tire model. The vehicle dynamics consider three degrees of freedom; longitudinal, lateral, and yaw; while the tire model is based on Pacejka’s detailed model. The integrated nonlinear vehicle model is implemented and validated on MATLAB/SIMULINK environment.
Since the vehicle model is highly nonlinear, different linearized models are obtained at different operating points. Since the vehicle has a wide operating range, a gap metric analysis is performed to determine the appropriate operating points at which the linearization process should take place. Therefore, multiple multi-input multi-output linear models are obtained to fully describe the vehicle’s nonlinear dynamics over the entire operating range.
Two control approaches are then proposed to achieve path tracking of the vehicle; linear quadratic regulator (LQR), and linear model predictive controller (LMPC). At first, a linear quadratic regulator is designed, then a better technique is implemented using the model predictive control approach. The designed MPC is based on Laguerre networks and aims to produce a smooth optimal control signal of the steering angle and the angular velocity. Furthermore, MPC design takes the steering angle and angular velocity physical constraints into consideration. To apply these controllers, their parameters for each linear model are designed and calculated offline. To combine these linear controllers, a fuzzy logic system is introduced to perform the online switching between the linearized controllers.
To test the capabilities of the proposed controllers, a path planning module is presented. Using this module, two paths with different degrees of complexities are generated. The first is a simple straight path that contains twolane-changing curves in the direction of motion. The second is a U-turn shape with different radii. Simulation results show the effectiveness of the proposed controllers. These controllers significantly improve the vehicle performance with less computational effort. The simulation results, also, prove the perfect tracking performance of the reference paths. For further investigations of the performance of the proposed controllers, a car following scheme is introduced. This scheme connects two car models; a car with a driver that generates the reference trajectory and an autonomous car with the designed tracking control methodology. Then, the tracking capabilities of the proposed controllers are tested and compared to the performance of a Linear Quadratic Gaussian (LQG) controller with a Kalman filter. The results of this study show the capability of the proposed integrated model predictive controllers to track different desired paths.