Street-crossing pedestrian trajectory prediction method based on SFM-LSTM neural network model

Abstract

The invention relates to a street-crossing pedestrian trajectory prediction method based on an SFM-LSTM neural network model. The method comprises the following steps: step 1, obtaining motion state information, individual feature information and pedestrian-vehicle interaction scene information of street-crossing pedestrians; 2, performing data preprocessing and data enhancement, and establishing a pedestrian trajectory data set; step 3, establishing and training an LSTM neural network model; 4, obtaining a prediction track of the street crossing pedestrian through the trained LSTM neural network model; 5, performing parameter calibration on the social force model by adopting a maximum likelihood estimation method; 6, correcting the prediction trajectory according to the social force model, and outputting the optimal prediction trajectory of the pedestrian crossing the street; and step 7, broadcasting the optimal prediction trajectory to nearby vehicles to assist intelligent network connection vehicles to make decisions, and compared with the prior art, the method has the advantages of improving the safety of pedestrian crossing, reducing the delay rate of the vehicles, improving the traffic capacity of roads and the like.

Description

technical field [0001] The invention relates to the field of intelligent network-connected vehicle-road coordination, in particular to a method for predicting the trajectory of pedestrians crossing the street based on the SFM-LSTM neural network model. Background technique [0002] In recent years, with the rapid development of intelligent networked vehicle technology, more and more models have different degrees of automatic driving capabilities. Lane Keeping Assistance (Lane Keeping Assistance) in Advanced Driver Assistant System (ADAS) System, LKAS), forward collision warning (Forward CollisionWarning System, FCWS), adaptive cruise control (Adaptive Cruise Control, ACC) and other functions have been installed on some mid-to-high-end brand models and become their standard equipment. [0003] Pedestrian safety protection is an important factor that ICVs must consider, but relying solely on single-vehicle intelligence to ensure pedestrian safety consumes a lot of on-board com...

AI Technical Summary

Problems solved by technology

However, the movement of pedestrians has a high degree of flexibility. The above-mentioned processing methods are either too conservative, making the driving process unnatural, affecting the traffic efficiency and the driving experience of the passengers on the vehicle, or insufficient analysis of pedestrian walking behavior, making accurate Therefore, the degree of protection for pedestrians is not enough, and the consequences are unimaginable in severe cases

[0005] In the pedestrian trajectory prediction method, the social force model (Social Force Model, SFM) is a method of pedestrian micro-dynamics. In a specific scene, given the initial conditions, the social force model can be used to generate a trajectory that conforms to the pedestrian walking law , using these generated trajectories as predicted values ​​is also a way of predicting pedestrian trajectories. Using the deep learning long-short-term memory network (LSTM) model to predict pedestrian trajectories, although the prediction effect is good, but only considering a single pedestrian in isolation The trajectory prediction of the target pedestrian does not consider the impact of vehicles and surrounding pedestrians on the target pedestrian

Images