| Intelligent vehicle is one of the future direction of automobile development.The control of intelligent vehicle contains longitudinal control and lateral control.Using traffic information around the vehicle from on-board sensors and vehicle motion states,intelligent vehicle longitudinal control calculates the expected longitudinal motion states,and then follows the target motion states by means of driving and braking systems.Compared with lateral control,longitudinal control has various traffic scenarios,high nonlinearity,most control methods in general have large irregularity and great differences with real drivers’ behavior.Therefore,longitudinal control of intelligent vehicle has become a difficult and hot spot internationally.Because of the randomness of road traffic behavior,the traffic environment of intelligent vehicle is very complicated with too many scenarios.Besides,intelligent vehicle longitudinal control operating mechanism is complicated,both driving and braking systems are needed,and actual drivers often control a car by coasting,with no throttle or braking control applied.In addition,the intelligent vehicle can also be affected by many disturbances,such as the differences of landforms,the change of the grade of road,the varying vehicle mass and load,variance of the air resistance and road attachment condition.All these factors lead to the complexity of intelligent vehicle longitudinal control.At present,most of the existing longitudinal control strategies designs control algorithms case by case.The actual vehicle complex driving scenarios are divided into several typical scenarios,applying different control algorithms for each typical scenario,and switching logic is used for switching between different algorithms.There is only one algorithm in action at the same time.Such methods cause control unstable because of the state discontinuity of the algorithms when the algorithms change,and the behavior of the intelligent vehicle is different from real drivers.Too many algorithms makes switching logic complex,which leads to heavy workload of calibration,and thus increases the cost of system development.According to the research status of the intelligent vehicle longitudinal control,a longitudinal control method capable of handling various driving scenarios based on driver optimal preview acceleration unified decision-making model is explored in this paper.With target longitudinal acceleration being considered as the decision intermediate variable,driver’s control behavior is divided into two parts,which are unified preview decision model for various traffic conditions and unified following-correction model for vehicle control.In unified preview decision model,the range of the desired acceleration is computed first and then discretized,and the preview acceleration set is obtained.Combined with the current motion states of vehicle,all the preview acceleration in the set is used to predict the future trajectory of vehicle.Firstly,the future trajectory is screened by the criterion of safety and compliance,then the optimal preview acceleration is decided by efficiency indicator and maneuverability indicator.This optimal preview acceleration is taken as the expected acceleration.In the unified following-correction model,a PID control combined with feedforward control is used to follow target acceleration.This control represents the basic driving law of real driver.In order to adapt to different vehicle control modes and system nonlinearity,according to the internal model principle in the forward channel an inverse system correction method is proposed to realize unified “1” system of vehicle response,so a single PID loop can be good enough to follow the desired acceleration.One single model can adapt to all control modes and driving environment.Virtual verification of intelligent vehicle based on driving simulator is widely applied at present,so a new generation integration platform for driving simulator to realise embedded verification of intelligent vehicle on driving simulator is proposed.The main contents of this paper are as follows:Firstly,research on unified preview decision model of vehicle longitudinal control.Apart from most decision making methods based on time headway,based on the original unidriver model of our research group,the optimal preview acceleration of vehicle is determined through four indexes of safety,compliance,legality,work efficiency and easy handling.A safety distance model based on driver’s preview behavior is proposed and a method for judging the safety by collision detection with predicted lane is proposed for the purpose of safety judgment.This model can get optimal preview longitudinal acceleration in various complicated traffic scenarios.Secondly,research on unified following-correction model of vehicle longitudinal control.The optimal preview acceleration is transferred into the control of the vehicle driving and braking system by the unified following-correction model.Based on the study of the control characteristics of real driver,a closed loop PID correction with feedforward to describe the basic driving law of real driver is proposed in this paper.Using inverse model of the corresponding low order vehicle longitudinal dynamics model of as forward channel,the characteristics of the plant are corrected into ideal “1” system at low frequency,which describes drivers’ understanding of different vehicle’s performance and makes the response of the driving and braking system identical at the same time,so a single PID closed loop can be able to achieve good following results,which can eliminate the assault impact of multiple PID controllers due to controller switching.An anthropomorphic phase splitting logic is proposed to solve the switching problem between throttle control and brake control.The effectiveness of the proposed method is verified with offline simulation with vehicle model developed by our research group.Thirdly,research and development of a new generation integration platform for driving simulator.Providing a realistic simulation environment and the ability of driver-in-the-loop simulation,virtual verification for intelligent vehicle longitudinal control system based on driving simulator has become the trend of the research of intelligent vehicle,which can effectively save time and money and ensure safety at the same time.In view of the fact that the simulator integration technology can not meet the requirement of intelligent vehicle verification,an integration technique for distributed real-time system is proposed in this paper,in which the communication,timer trigger,system configuration,real-time solver,command system and script automation technology are explored.In a distributed system consists of common PCs and multiple LANs,data communication,timing and messages are separated to achieve millisecond level real-time integration.Through multicast,a Simulink like MIMO topology is implemented between distributed real-time applications.The driving simulator is integrated by this technology,and driving simulator can be seamlessly integrated with the intelligent driving systems and the d SPACE hardware and software toolchain.Taking longitudinal control as an example,a virtual verification system based on driving simulator and a vehicle test platform are built.Finally,experimental research on integrated longitudinal control of intelligent vehicle based on driving simulator.Systematic research of intelligent vehicle longitudinal control evaluation criteria and test conditions are made for evaluating control method,and test cases and intelligent vehicle longitudinal integrated control evaluation method based on power,economy,driveability and anthropomorphism of the control behavior are proposed.All the test cases are realized based on traffic simulator of driving simulator,and virtual experimental verification of the control strategy is carried out on driving simulator.The results show the effectiveness of the proposed control strategy and the system integration method. |
PDF Full Text Request