Research On The Hybrid Mechanism And Rule Based Modeling For The Comprehensive Decision Of Intelligent Vehicles' Direction And Speed

The intelligent vehicle has become the tide and trend of the automobile development in today's world.After years of study and research,intelligent driving assistant technology for simple driving condition is maturing.Advanced driver assistance systems like lane departure warning,adaptive cruise control and lane keeping assist have been widely used in the automobile industry.Besides,partial automatic driving and conditional automatic driving have entered the testing period.Now researchers are working on highly automated intelligent vehicles,and one of the intractable problems is the comprehensive decision model of direction and speed for the complex traffic environment.Two types of factors must be considered when the driver is making decisions on how to control the direction and speed of the vehicle.One is the factors of continuous value,such as driving work efficiency and handiness,etc.The other is the factors of binary logic,such as driving safety and legality,etc.Most of the existing models are mainly based on control theory and the mechanism modeling method.The discontinuous factors and the continuous factors are combined into the comprehensive objective function to solve the optimization problem,which makes it difficult to determine the weighting coefficients,and easily leads to the divergence of the optimization process and dangerous situations such as the vehicle moving out of the road boundary under an emergency turn.To solve the above-mentioned problem,a modeling approach of hybrid decision-making based on mechanism and rule is proposed.On the basis of the optimal preview acceleration model of the integrated control of the direction and speed,the safety and legitimacy are modeled based on rule and used as constrains to reduce the feasible region of the preview longitudinal acceleration and preview lateral acceleration.The driving work efficiency,handiness and so on are modeled based on mechanism and used as the objective function to solve the comprehensive optimization problem.The same practice has not yet been seen in the current literature and this thesis focuses on the following aspects:Firstly,the feasible region reduction method of preview acceleration is studied.In this thesis,we investigate the driving safety of the preview longitudinal acceleration and the preview lateral acceleration,which are based on the constraints of the road accessible area and obstacles.Also,we study the legality of speed regulation,lane lines,traffic lights and stop lines to reduce the feasible region of preview acceleration and ensure the optimal preview acceleration in line with the requirements of vehicle safety and traffic rules.We study the method for judging the relative position between the vehicle contour and the road accessible area,obstacles and lane lines and the phase separation method for judging the legality constranined by traffic signal lamp and stop line.The simulation and test results show that the proposed method has effectively solved the problems caused by the rule factors involved in the comprehensive optimization,such as difficult to determine the weighting coefficients and the divergence of optimization process.Secondly,the comprehensive evaluation method of preview acceleration is studied.In the feasible region,which is restricted by the safety judgment and the legitimacy judgment,the optimal preview acceleration is optimized by the comprehensive evaluation of the evaluation indexes such as driving work efficiency,handiness and so on.The original model mainly depends on the lateral safety index to drive the vehicle close to the lane center line.However,the lateral safety index mainly changes in the area near the lane line or near the road boundary,and the rate of changing in the vicinity of the lane center line is very small,which lead to poor convergence vehicle track,driving not stable along the lane center line,and showing a continuous swing.Modeling based on safety rules will further exacerbate this problem.Thus,we propose a new index for evaluating the lateral following performance,which is used to describe the psychological characteristics of the driver's expectation that the vehicle is going close to the central line of the lane.The simulation and test results show that the proposed method effectively solves the problem of poor convergence.Thirdly,the dynamic correction method of nonlinear dynamics is studied.In the classical optimal preview acceleration model,the low order equivalent inverse system model of the controlled vehicle is used to open-loop correction.The consistency between the actual acceleration and the optimal preview acceleration can be guaranteed by accurate calibration of the inverse model.Because of the dynamics of the vehicle,there is the central area and the large slip zone in addition to the linear region,presenting strong non-linearity.In order to obtain the parameters of the inverse model accurately,it is usually necessary to do calibration experiments on a large number of operating points which are composed of vehicle speed,acceleration and road surface conditions.To solve this problem,this thesis proposes a new method based on the principle of compound correction control.The open-loop correction is as the main body on behalf of the driver's mastery of vehicle dynamics,a closed-loop correction link is added on the basis of this,which can be used to describe the compensation correction ability.The simulation and test results show that this method can make the model robust,and the calibration difficulty is reduced to a great extent.Finally,the hybrid mechanism and rule based modelof comprehensive decision model of direction and speed is established in the Simulink environment.The C-class demo vehicle model of vehicle dynamics simulation software Car Sim is used as the control object to carry out the simulation of the driver-vehicle-road closed-loop system such as the double lane changing section,the right angle turning section,the lemniscate section,the 3D cycle track,the cruising,overtaking,and the traffic signal lights.In addition,typical ACC conditions are tested with the driving simulator as a test platform.Simulation and test results show: according to the proposed modeling method,the weighting coefficient of each evaluation index can be simply determined and well adapted to the above conditions,and there is no unsafe or violation of traffic rules.It effectively solves the problem that the optimization process of the existing model is easy to diverge.The vehicle can travel steadily along the center line of the lane after lane-change or curve passing.There was no undesirable phenomenon of sustained oscillation,and it solves the problem of bad convergence of the original model.By simply calibrating the inverse model of the controlled vehicle,we can ensure that the actual acceleration has a high consistency with the optimal preview acceleration,and effectively solve the calibration problem.