Research On Key Improvement Of Driving Simulator Fidelity

Instead of paving ground tests,it is a mainstream of developing automotive products using all-digital or co-simulation,which combines model and physical parts together.Vehicle performance mainly includes system performance under steady state condition(steady state performance)and dynamic process under transient condition(dynamic quality).In the stage of product pre-development,pure digital prototype simulation can be used to realize the steady state performance integration and matching.But the dynamic quality could only be tuned after the demo car being produced.As the vehicle layout and system structure have been settled,the range is limited when matching the dynamic response based on subjective evaluation results,increasing the risk of development.The technical route was proposed internationally that using the driving simulator for subjective evaluation.It combined the virtual paving ground and subjective evaluation,so that the subjective evaluation results can always be controlled in the whole product development cycle and improved the development efficiency.Driving simulator has been widely applied in the development of advanced electronic control system(ECU),human-vehicle-environment closed-loop system,intelligent vehicle verification and other fields,becoming a novel trends in the automotive industry.The objective performance of the vehicle dynamics model adopted by the simulator is consistent with the real vehicle,but the driving feeling is very different from real vehicle,which hinders the application of the simulator in the automotive industry.At present,scholars all over the world are engaged in the research of simulator fidelity,which has become a hot topic in recent international Driving Simulator Conference.The driving simulator unlifelike phenomenon mainly include: maneuverability incorrect,unreal road sense,unreal speed perspective,image flicker,simulator sickness and other inconsistent phenomenon with the real vehicle.The phenomenon reflect in braking drift,stand-still drift,parking torque inaccuracy,without on-center feeling,poor stability,poor linear stability,unreal road sense,lower speed perspective,driving like "sailing".After analyzing the reason of the low fidelity,it is found that the vehicle dynamics model adopted by the simulator is a performance model based on objective evaluation,which lacks the modeling of the key aspects of the vehicle's dynamic process and can't simulate critical conditions such as low-speed or parking which affect the fidelity of maneuverability.Besides that,because of the limited compensation capability of the existing simulator correction algorithm which cause sickness and image flicker,reduce the immersive sensation used by the simulator.In addition,Simulator virtual scenes are mostly non-horizontal road and lack of road roughness which cause vehicle vibration is an important part of the driver's sensory for road and speed perception.From the above reasons,there are some limitations to use the simulator for subjective evaluation.Based on the important role of the simulator in the development of automobile products,the hard work was attempted to promote the fidelity without upgrading hardware by improving the vehicle dynamics model,simulator integration and driving road environment in this paper.Firstly,the simulation accuracy and condition adaptation of vehicle dynamics model are improved,and the accuracy of simulator signal source is guaranteed.A steering-wheel dynamic model is builted,and the steering system model is changed from a steady state model for steering simulation to a dynamic model for stability and returnability simulation and improved the maneuverability and extended simulation conditions.Secondly,a model based prediction method is proposed,which is to solve the problems that high frequency shake and instantaneous signal do not conformity with the vehicle motion caused by the simple compensation of simulator large delay.At last,a road roughness model is built,proving a realistic driving environment,which improves the perceptions of speed and road feeling.Concretely,this paper is carried out by following contents:Firstly,for the maneuverability modeling,in the paper,a dynamic steering-wheel modeling strategy under full simulation condition was researched.Currently,almost of the commercial dynamics models are steady-state model based on objective performance evaluation.The steering system adopts forward kinematics and inverse mechanics modeling,in which the mechanical relationship between the left and right wheels in the steering mechanism is isolated by displacement constraints.The tire model based on slip rate cannot realize the simulation of low speed,zero speed and other extreme conditions,and does not meet the requirements of simulator whole working condition simulation.The tire model uses the relaxation length method to express the elastic effect of the carcass,resulting in the absence of the elasticity of the steering system.These models can simulate the linearization characteristics of the vehicle in the steady state condition correctly,but can't accurately simulate the transient response and the ability to resist the external disturbance,causing the steering too sensitive,and easy to lose stability at high speed and steering torque is incorrectly at low speed,so inadequacy used it for the subjective evaluation on the driving simulator.Base on this,the paper established a complete dynamic model,integrated Ackerman steering mechanism and steering components,a dynamic wheel model considering carcass elasticity and the static-dynamic friction model are established in this paper.Compared to the existing steering model,the transient process and transient force simulation is realized,and the dynamic stability for external disturbance are described accurately.In addition,the model realized the critical conditions simulation such as ramp stop,brake start,parking steering and so on,improves the fidelity of maneuverability and extended simulation conditions and inhanced the fidelity from the singal.Secondly,for the compensation of the simulator large delay,a method of Model based Prediction Correction(MBPC)was proposed in the paper.When the traditional method of first order or second order differential prediction is applied to the large delay system,it will cause the prediction error and oscillation.At the same time,the use of signal-based prediction theory,without considering the specific physical meaning of the signal,will destroy the correlation between the different signals,resulting in instantaneous prediction values that cannot match the characteristics of the system.Aiming at the problem of transmission and response delay and simulator system matching,this paper proposes an comprehensive correction method based on the kinematics of the vehicle body.By considering the unique movement characteristics of the vehicle,so as to ensure the predicted instantaneous vehicle state meets the characteristics of vehicle dynamics.In order to solve the problem of delay in different systems,the paper adopted the synchronizing prediction the vehicle states at the future time,which ensures the consistency of signal sources of different systems.Aiming at the driving simulator in which the screen is fixed,the paper also studied the correction of the visual system and the motion system,avoiding the conflict between visual and somatosensory.The proposed prediction method improved the consistency between the simulator and the real vehicle response.In addition,from the point of enhancing the road sense and speed sense,the model of the digital road was researched in the paper.Traditional virtual scenes are generally based on the purpose of visual information based on the horizontal road,non-level road,lacking of road high-frequency excitation,so the driver can not perceive the high-frequency excitation from the road to the chassis,seat and steering wheel,resulting in the lack of some sense,affecting the driver's judgment on the state of the vehicle.The paper redefines the physical characteristics of the road and generates 3D-road roughness data with fixed area.Aiming at the requirement of real-time simulation of large scene in simulator,the paper studied the modeling method of non-level road,real-time detection algorithm and smooth transition method;Based on this,the paper also studied the smooth stitching method and fast interpolation algorithm for the road surface roughness.The digital road model improves the road feeling and velocity fidelity of the simulator.Finally,integrate the established model into the driving simulator for subjective and objective evaluation.Aiming at the inconsistency of the calculation cycle in different system of the simulator,a data processing model is established to avoid the frequency aliasing or step effect between different systems.And according to the characteristics of simulator data application,the data transmission path is optimized.The vehicle dynamics model is integrated and compared with the field test to verify the accuracy of the dynamic model.The above model is integrated into the simulator for subjective and objective simulation verification,and the simulation results show that the adaptability of the dynamics model and the accuracy of the steering maneuverability by driver are verified from the test of the free steering wheel,the parking steering,the small steering input and the steering handiness and so on.The accuracy of the road roughness model is verified by the driving test on the different grades roughness road,which improves the fidelity of the perception of road and velocity feeling.The effectiveness of the proposed delay correction model is verified by driving with a fixed path("S road")and objective measurement.In summary,the simulation and subjective evaluation results show that the model established in this paper can enhance the fidelity of the simulator effectively and can be applied to the subjective evaluation in a certain extent.