Method for generating trajectory prediction model of vulnerable road user and corresponding prediction method

Abstract

The invention discloses a method for generating a trajectory prediction model of a vulnerable road user and a corresponding prediction method. The method comprises the following steps: step S1, obtaining a training set, wherein the training set comprises marked N + M frames of images which are continuous in time; step S2, screening training samples from continuous N + M frames of images in the training set according to the tracking ID serial number of each target true value, wherein the training samples of the same target with the same tracking ID serial number are in one group; step S3, extracting space-time multi-clue features and acquiring a training input vector Xjt + k and a training output vector Yjt + k; and step S4, generating a model, inputting the training input vector Xjt + k and the training output vector Yjt + k into a gating cycle unit neural network adopting an adaptive two-parameter activation function, and performing model training by adopting a pre-designed loss function to generate a vulnerable road user trajectory prediction model.

Description

Technical field [0001] The present invention relates to the field of automatic driving, in particular to a method for generating a trajectory prediction model for vulnerable road users and a corresponding prediction method. Background technique [0002] Pedestrians and cyclists (including cyclists, electric bicycles and motorcycles) in traffic scenes are vulnerable road users (Vulnerable Road Users, abbreviated as VRU). Predicting the trajectory prediction of vulnerable road users around smart vehicles is one of the key technologies of smart transportation, and it is the basis for smart vehicles to perform trajectory planning and movement obstacle avoidance. [0003] The main research object of existing VRU trajectory prediction methods is pedestrians. Moreover, traditional pedestrian trajectory prediction methods only focus on the movement time clues of position changes, ignoring appearance characteristics. In other words, existing trajectory prediction models or methods do not ...

AI Technical Summary

Problems solved by technology

Moreover, traditional pedestrian trajectory prediction methods only focus on the motion time clue of position change, ignoring appearance features

That is to say, the existing trajectory prediction models or methods do not consider the movement characteristics of different types of VRUs. For example, because the riding tool of the lower body of the cyclist is a rigid body, the randomness is less than that of the pedestrian, and the movement speed is faster.

In addition, the existing trajectory prediction models or methods do not consider the interconnection between VRU targets, and do not make full use of the temporal motion features and rectangular frame appearance features of VRU targets, as well as the shared contextual image features between different VRU targets.

Therefore, the accuracy of the existing VRU trajectory prediction model or method is poor, and it is not suitable for the trajectory prediction of vulnerable road users in complex and changeable scenarios

Images