Design And Experimental Validation Of Longitudinal Controller Of Connected Vehicle Using Model Predictive Control

With the development of the world economy, vehicle has been widely used by people and more and more people start to drive. But the increased number of people's driving result in more accidents in the world. Besides having people to have a good driving habits and safety consciousness, autonomous vehicle is considered to be an effective way to eliminate the damages of accidents caused by people's driving, because the safety, efficient, and convenient of autonomous has been proved by a lot of tests. Now some advanced driver assistance system such as Adaptive Cruise Control, Lane Departure Warning, and Blind Point Detection have been used on passenger vehicles. But these functions are far away to the autonomous vehicle, to find another milestone technology is essential to the autonomous vehicle in the future.The concept of connected vehicle was introduced in this work where the vehicles have ability to share the dynamic information of vehicles with the surrounding vehicles. Control system which is based on Model Predictive Control was designed to maintain the desired relative dynamics between the leading vehicle and the following vehicle in vehicle following maneuver. Compared to the Conventional Model Predictive Control, the Laguerre polynomials was introduced to approximate the control signal. Then the selection of different parameters of Laguerre polynomials yield to different characteristics of the closed-loop system. The control system was implemented in the platform which is finished by Federal Highway Administration, the functionality of the designed controller was proved by the testing results.First, the kinematic relationship between the leading vehicle and the following vehicle was described in vehicle automatic following scenario. The following vehicle is always trying to maintain a desired distance and relative speed between the two vehicles. The generating process of augmented model which is specific used for the design of MPC was introduced. Combined with the model which can show the kinematic relationship between the leading vehicle and following vehicle, the augment model was developed. One benefit of the augmented model is that there is always a integrator in the model so that the state error can be eliminated.After the development of the augmented model, the basic principle of MPC was presented. Compared to the conventional MPC, Laguerre polynomials was used to approximate the control signals. In order to show how the Laguerre polynomials approximate the signal, the results of approximation to the impulse response to the system were compared. Then the suitable parameters of Laguerre polynomials were determined. When Laguerre polynomials was applied to the design of MPC, the special characteristics of Laguerre polynomials such as orthogonality and exponential decay yield a simpler expression and an easier solution. Moreover, the control performance can be tuned by different parameters of Laguerre polynomials. After the parameters and the weighting matrix were determined by the simulation results, the control system was simulated in Matlab and the results show that the controller worked fairly well.Finally, the designed controller was implemented on the platform in Federal Highway Administration. The components of the platform include the positioning system (PinPoint), communication system (DSRC), TORC speed controller, and MicroAutobox. PinPoint is responsible for supplying the positioning information and dynamic information; DSRC enable the vehicle have the ability to share information with surrounding vehicles; TORC speed controller can implement the speed command which is generated from the higher level controller; the whole control system is implemented in MicroAutobox. The testing results show that the control system can successfully maintain the desired distance and desired relative speed between the two vehicles. Meanwhile, some drawbacks can be found through the testing results and solutions were proposed. The solutions which are supposed to improve the control performance in the future were all verified by simulations.