LSTM model-based vulnerable traffic participant trajectory prediction method

Abstract

The invention relates to a vulnerable traffic participant trajectory prediction method based on an LSTM model, and the method is characterized in that the method comprises the following steps: 1, selecting vulnerable traffic participants and vehicles in a zebra crossing region under the condition of mixed driving of people and vehicles, and carrying out the early-stage investigation; step 2, acquiring motion state information, individual feature information and interaction scene information of street-crossing vulnerable traffic participants; 3, establishing an LSTM (Long Short Term Memory) model, and training the LSTM model; and step 4, performing trajectory prediction on the street-crossing vulnerable traffic participants through the trained LSTM model, and obtaining predicted trajectories of the street-crossing vulnerable traffic participants within a first preset duration in the future, compared with the prior art, the method has the advantages of improving the street-crossing safety of the vulnerable traffic participants, improving the traffic capacity of the road and the like.

Description

technical field [0001] The invention relates to the field of automatic driving decision algorithms, in particular to a trajectory prediction method for vulnerable traffic participants based on an LSTM model. Background technique [0002] In recent years, with the increasing number of cars in the world, traffic accidents between cars and Vulnerable Traffic Participants (Vulnerable Traffic Participants, VTPs) occur frequently. Key issues such as how to ensure the travel safety of vulnerable traffic participants and reduce the rate of automobile traffic accidents need to be resolved urgently. [0003] The existing self-driving car protection system for pedestrians, non-motor vehicles and motor vehicles is mainly based on the target detection algorithm. Through the detection and identification of traffic participants, if there is a risk of collision, the self-driving car will carry out early warning and avoidance. Touch control. Especially for vulnerable traffic participants s...

AI Technical Summary

Problems solved by technology

The simplest models include constant velocity (Constant Velocity, CV) model and constant acceleration (Constant Acceleration, CA) model. It is suitable for short-term trajectory prediction, but when the external environment changes (such as obstacles in front, vehicle deceleration in front, etc.), this method cannot accurately predict the future trajectory of vulnerable traffic participants

There are also social force models used to predict the trajectories of vulnerable traffic participants. Although the prediction effect is good, it can only simulate the immediate movement responses of vulnerable traffic participants. It cannot consider long-term dependent information and adapt to complex movements like neural networks. scenarios, which will also reduce the accuracy of trajectory prediction for vulnerable traffic participants

Images