Research On Environment Perception Technology For Intelligent Vehicle Based On Multi-Source Information Fusion

As a national strategy of China,the policy of developing intelligent automobiles has become an essential breakthrough in promoting the transformation of the national economic strategy and achieving the supply side reform.Among the technologies of autonomous vehicles,the environmental perception is a key bottleneck and will severely affect the application of driverless cars.In order to provide comprehensive information for the control systems and decision-making systems of unmanned vehicles,it is necessary to accurately perceive the surrounding environment in real time for vehicles.Therefore,the means to solve this issue primarily involve in efficient on-board sensing devices or systems,e.g.,the system with integrating various sensing units,which will product great practical significance and engineering value.The research of this paper includes the parts as follows.1)Architecture of the intelligent automotive environment sensing system based on multi-source information fusion.By analyzing the basic theory of multi-source information fusion and the developing process of intelligent automotive environment perception,this paper summarizes the function framework of information fusion in multi-sensing collaborative perception of single cars and multi-sensing collaborative position of multi-vehicles.In addition,this paper analyzes the main features of input data in the level of target fusion,i.e.,multi-source,heterogeneous and variable scales,and sum up the five key techniques in the perception method,which will lay a theoretical foundation for perceive methods in the information fusion of single cars or multiple cars in future.2)Perception method of multi-sensor information fusion of single cars.In terms of the current issue that the single sensor cannot effectively identify the target vehicles ahead in the field of engineering application for intelligent automobiles,this paper puts forward a target identification scheme based on the information fusion of radars and cameras.The fusion method involves the hardware framework that uses the information of radar basically and camera auxiliarily,introduces Markova distance to match observation of target sequence,and establishes systematic observation model and state model base on the joint probabilistic data association(JPDA)method for data fusion.Finally,the effectiveness of the algorithm is verified by software simulation and real vehicles test.3)Perception method of multi-sensing collaborative position of multi-vehicles.The paper presents the overall framework of multi-sensing collaborative position of multi-vehicles and clearly shows the necessary information and the primary functional modules.Specifically,the location estimation algorithm based on extended Kalman filter has been proposed by using the observed information of the vehicles.Besides,this paper also proposes two sets of vehicle cooperative position algorithm involving the optimal fusion estimation of single packets and the suboptimal fusion estimation with the historical information.Meanwhile,the paper proposes a method of target correlation between the observation information from radars based on joint probabilistic data association and the information of vehicle targets accessed by communication technology by measuring the similarity between observation values of Markov distance.The results of simulation show that the algorithm considering the historical information has excellent adaptability to the poor communication condition,and the effectiveness of the target correlation algorithm has been verified.4)Analysis of impact of communication indexes on the application of intelligent automotive environmental perception.As the current main communication technologies of car-to-car(e.g.,DSRC,LTE)cannot guarantee to maintain an ideal state under all working conditions,this paper focuses on the scenarios of linear motion and lane changing.Through the analysis of the application of intelligent automotive environmental perception from different communication indexes,such as packet loss rate,latency and the limitation of communication distance,this paper provides the reference standard for the design of information fusion environmental sensing method on car-to-car communication.The innovation points of this paper include the following components:1)Construct the fusion system framework of intelligent automotive environment awareness,and the framework consists of different sensor combinations.2)Propose a perceive method based on the information fusion in multi-sensing of radars and cameras for single cars.The perceive method has proved to be more excellent in recognition accuracy than the method using merely one sensor.3)Put forward a high-precision perceive method based on multi-sensing collaborative position of multi-vehicles.Considering the economic factors,this method increases the application prospect of engineering in future.4)Analyzing the influence of communication quality on the engineering problem of the application of intelligent automotive environment awareness.