The Research On Drivable Region Modeling For Intelligent Vehicle And Display Technique Of Driver Assistance Information

The intelligent vehicle realizes the semantic level understanding of the surrounding environment by simulating the real driver's insight into the driving road conditions,and using the environmental perception system and information fusion technology(referred to as environmental perception technology).At the same time,the anthropomorphic vehicle behavior decision process is provided more comprehensive information support.This paper relies on the sub-project “Research and Demonstration of Complex Road Environment Synergy and Target Tracking” of the National Ministry of Science and Technology project “Key Technology Research and Demonstration Operation of Electric Autonomous Vehicles”,and proposes the concept of intelligent vehicle drivable region modeling,aiming at the core problem in the intelligent vehicle environmental perception technology.The concept of drivable regional modeling is the compact expression model of the current vehicle environment,which is established by combining the prior knowledge with the driving state information acquired by the on-board sensor.Vehicle drivable regional model also deepens the understanding of the intelligent vehicle to the environment and enhances the driving safety.The main contents of this paper are as follows:(1)Research on prediction and behavior recognition algorithms for vehiclesFirstly,in order to predict and identify the driving intentions of other vehicles on the road,this paper uses the unscented Kalman filter to track and predict the vehicle motion state based on the vehicle motion model.Secondly,based on the hidden Markov model,the vehicle behavior recognition model is established,which realizes the prediction and probabilistic estimation of the vehicle's future behavior intention.Finally,offline training and functional testing of the model are carried out using NGSIM(Next Generation Simulation)traffic dataset.(2)Algorithm for generating and updating dynamic probabilistic drivable mapIn order to visualize the drivable region of the intelligent vehicle,this section firstly constructs the dynamic probabilistic drivable map,and uses the lane estimation to mesh the map.Furthermore,based on vehicle behavior recognition and prediction methods,the map's ability to understand other vehicle behaviors in the road environment is enhanced.Finally,through the layered fusion processing of vehicle information,lane environment information,road vehicle information,and road vehicle prediction information,the solution process of each cell's drivable probabilistic is described.(3)Design of driver assistance information based on dynamic probabilistic drivable mapIn order to improve the human-machine interaction of dynamic probabilistic drivable map,in this paper,firstly,proposes the method of establishing the behavioral decision cost function based on the requirements of driving assistance system,and uses the dynamic programming algorithm to design the decision model of the minimum cost driving path of the vehicle.Furthermore,the software system framework including the above decision algorithm and dynamic probabilistic drivable map is constructed through modularity.Finally,the humanmachine interaction program is designed and developed to meet the interaction requirements between the driver and the system.(4)NGSIM dataset for model training and driving assistance decision system function verificationFirstly,in order to train the behavior recognition model based on hidden Markov model,this paper using the NGSIM traffic dataset,and preprocesses and extracts the trajectory effective data.Secondly,in order to verify the validity and feasibility of the model and algorithm established in this paper,two typical scenarios of intelligent vehicle driving are designed based on Qt Creator?.The feasibility of the algorithm is verified by three kinds of simulation tests.Finally,using the self-driving vehicle test platform,the actual test of the above typical conditions were carried out.The results show that the driving assistance information display system designed in this paper can accurately and effectively express the complex road environment,and provide reasonable driving advice according to the specific environment,which has good robustness.