The Research On The Correction Module Of Driver Direction Control Model

Transportation Safety has been a social problem with the popularization ofautomobile and the development of automobile technology. The developmentof intelligent vehicles tra?c and the automobile active safety have been paidmore and more attention. Throughout the research, as the leading factors ofthe automobile active safety, the vehicle handling stability and the driver's be-havior property play an important role in it. At the same time, these two onesare interactively connected within the driver-vehicle close loop system. It hasbeen proved that considering the relationship among the driver, vehicle andthe road as a whole and to investigating the vehicle handling stability from thesystematic aspect is on the right tack, both of which had been made possible bythe development of the driver model, which has importance significance on thestudy of the closed-loop evaluation of vehicle handling and stability and intel-ligent tra?c especially intelligent vehicles. With the development of advancedautomobile technology, they have changed vehicles mostly and promoted theperformance and safety of vehicles. We can use driver model to evaluate the ve-hicles'handling and stability comprehensively, so we can make the reformationof advanced automobile electronic technology and reduce the time and cost ofresearch and development.The purpose of this thesis is: Because the lateral dynamics of vehicles is stronglynonlinear, in the design process of the correction module of driver model itshould acquire equivalent linear model of vehicles and then design the correctionmodule according to the parameters. This thesis did parameter identificationby 1-order linear model before, so this thesis contrasts between 1-order modeland 2-order model, and think 1-order model can get better identification results.This thesis used least square method to make parameter identification and metsome problems, so this thesis use generalized least total variance fit methodto make identification. This thesis contrast between these two methods, andthink generalized least total variance fit method can get better results. Basedon these, this thesis uses Model Reference Adaptive Control method to design the correction module, this method doesn't need identification, and can reducegreatly reduce the workload of design.Firstly, based on the preview-follow theory we must firstly get the transferfunction of the lateral dynamics of vehicles. So this thesis need to make systemidentification for it. This thesis firstly use least square method to do it, anduse ARX model. This thesis make parameter identification at the di?erentvelocities, and fit the data. The controller chooses the proper parameters fromthe fitted data according to the current velocity. With the consideration ofdriver nerve delay, this method can acquire good control e?ect at relativelyhigh speed. There is vibrate with the follow curve in the lower speed. Withoutthe consideration of driver nerve delay, his method can acquire good controle?ect, and there isn't vibration with the follow curve in the lower speed. Thisthesis contrasts between 1-order model and 2-order model, and think 1-ordermodel can get better identification results.Secondly, this thesis also use generalized least total variance fit method to opti-mize the parameters which can reduce the error of identification. If this thesismake unit step response for the identified system and real system, this thesiscan find their stable values is equal, but the rising parts inosculate badly. Thisthesis use generalized least total variance fit method in the parameter identifi-cation which can optimizes the parameters according to the restrictions. Thisthesis make use of o?-line identification for this process which can improve thespeed of identification. This thesis make parameter identification at the di?er-ent velocities and di?erent angles of steering wheel, and form 3-D surfaces ofcontrol parameters. Controller will choose the proper parameters from the 3-Dsurfaces according to the velocity and the angle of steering wheel, so the con-troller use the parameters of equivalent linear system at any time. This thesiscan find this method can control the system with driver nerve delay, and theangles of steering wheel accord with the behavior of real drivers. This thesiscontrasts between 1-order model and 2-order model, and think 1-order modelcan get better identification results.Lastly, according to the adaptive control theory and preview-follow theory, thisthesis uses Model Reference Adaptive Control method to design the correction module. This thesis firstly introduces the basic theory of adaptive control, anduses Model Reference Adaptive Control PID Controller as the control-adjustpar. Lastly, this thesis makes simulations at relatively high speed and lowspeed, and makes contrast and analyst. This thesis find this method overcomesthe di?culty of parameter identification and disadvantage of great workloadand there is advantage of preview time. When you use this method, it meansyou don't need knowledge about controlled vehicles. Although this methodhas some limit, it is also a good road for research. At last, this thesis thinkself-tuning control can solve the problem which can't be solved completely now.In a conclusion, this thesis make many researches about the lateral dynamicsof vehicles and preview-follow theory, and using 3 di?erent methods to controlvehicles'direction. In these ways, 2 ways need to make parameter identification,1 way don't need to do it. The first identification method is simple but haslimited capacity, another way need complex identification but have powerfulcapacity. At the same time this thesis think 1-order model can get betteridentification results. Model Reference Adaptive Control method is a goodmethod, it means you don't need knowledge about controlled vehicles. So thisis a fascinating method, and we will continue to develop this method.