Trajectory prediction method for vehicles surrounding urban intersection

Abstract

The invention discloses a trajectory prediction method for vehicles surrounding an urban intersection. The method comprises the steps that 1) vehicle state information obtained by a sensor serves as input, and a crossing intention identification model and a give-way intention identification model are used to obtain a motion mode of a target vehicle; 2) after that the motion mode of the target vehicle is determined, a future driving trajectory of the target vehicle needs to be predicted, and for each motion mode, a corresponding acceleration prediction model is established to obtain the predicted acceleration; 3) after that the predicted acceleration of the target vehicle is obtained, a constant acceleration model is used to calculate the motion state of the vehicle in next step; and 4) in the practical using process, unscented Kalman filtering is combined to reduce an error of the prediction model. Thus, real and valid vehicle state data is provided for training and testing of the trajectory prediction model of the urban crossing.

Description

Technical field [0001] The invention relates to the field, in particular to a method for predicting the trajectory of vehicles around an urban intersection. Background technique [0002] As a future tool for improving the efficiency of urban road traffic and reducing road safety accidents, unmanned driving related technologies have been greatly developed in recent years, and many scientific research institutions and universities at home and abroad are still increasing their research and development efforts. When driving on actual roads, unmanned vehicles cannot make better driving decisions in time and accurately in an unknown environment with incomplete experience and knowledge like human drivers. An important reason why decision-making problems are difficult to solve is the insufficient accuracy of environmental element modeling (especially dynamic element modeling). For example, when driving at an urban intersection, the driver can often predict the movement patterns and futu...

AI Technical Summary

Problems solved by technology

[0005] For the prior art, because the trajectories of vehicles with different motion patterns present different rules, the prediction accuracy of the trajectory prediction model will be insufficient, and the future long-term driving trajectory of the vehicle cannot be accurately predicted, and it is not suitable for the complex urban intersections targeted by the present invention. For the long-term trajectory prediction problem, a vehicle trajectory prediction method around urban intersections is proposed

Images