Research on Autonomous Driving Perception and Recognition Technology at Intersection

Due to complex types of traffic participants and large randomness of traffic behaviors, intersection has become a typical scenario of urban traffic accident and a severe challenge for the application of autonomous driving urbanization. This paper extracts typical cross shaped intersection conflict scenarios through theoretical derivation in the first place. Next, to evaluate if single vehicle multi-sensor information fusion technology is reliable at intersections, a high mileage real road test is conducted. Then the potential risks of this scheme at intersection scene are analyzed. Finally, a technical concept of automatic driving perception combining V2I and multi sensor information fusion at intersections is proposed.