A Real-time Recognition Method of Surrounding Vehicle Behavior Based on Kinematics Prediction and Compensation Mechanism

Abstract

The invention discloses a real-time identification method of surrounding vehicle behavior based on kinematics prediction and compensation mechanism, including: typical behavior of surrounding vehicles and hot zone definition: classifying the typical behavior of surrounding vehicles, combining lane lines and road shoulder positions, Divide into 5 hotspot areas; offline training of hidden Markov vehicle behavior recognition model: establish discrete hidden Markov model DHMM for each maneuver class, and use EM algorithm to train each DHMM model to obtain the best DHMM model group; Establish a vehicle trajectory prediction kinematics model: The vehicle trajectory prediction kinematics model can predict the trajectory and generate the position sequence information of the vehicle for the next 3 time steps; online real-time recognition: the obtained vehicle history 5 time step position sequence and the predicted vehicle future 3 The time-step position sequence is transformed into an 8-time-step hotspot sequence, which is input to the trained hidden Markov vehicle behavior recognition model, and the surrounding target vehicle behavior is obtained by forward calculation and recognition.

Description

technical field [0001] The invention belongs to the technical field of intelligent driving, and in particular relates to a method for real-time recognition of surrounding vehicle behavior based on a kinematics prediction and compensation mechanism. Background technique [0002] Nowadays, whether it is advanced driver assistance systems or fully autonomous vehicles, scholars in various fields have aroused extensive research interests. There is no doubt that automotive intelligence has become one of the most important trends and trends in the development of the automotive industry. In addition, the 5G communication era is coming, and our country is one of the leaders leading this era. The V2X communication of 5G Internet of Vehicles based on D2D technology has an air interface delay of about 1ms and an end-to-end delay of milliseconds. Accurately obtain the status information of surrounding vehicles in real time under the scene. At this stage, the biggest challenge for us to...

AI Technical Summary

Problems solved by technology

The existing hot spot feature conversion method only uses the historical location information of surrounding vehicles, and the recognition is often completed after the behavior has been basically completed. At this time, the recognition result has little practical significance for intelligent vehicles, and lacks real-time recognition and applicability.

In addition, in the initial stage of the execution of a certain surrounding vehicle behavior, the recognition accuracy is generally low due to the incomplete historical information of the behavior in the observation sequence and the interference of the previous vehicle behavior information.

Images