Critical Example Mining for Vehicle Trajectory Prediction using Flow-based Generative Models

TL;DR

A critical example mining method that utilizes a data-driven approach to estimate the rareness of the trajectories and indicates that the mined critical examples include uncommon cases such as sudden brake and cancelled lane-change, which helps to better understand and improve the performance of prediction models.

Abstract

Precise trajectory prediction in complex driving scenarios is essential for autonomous vehicles. In practice, different driving scenarios present varying levels of difficulty for trajectory prediction models. However, most existing research focuses on the average precision of prediction results, while ignoring the underlying distribution of the input scenarios. This paper proposes a critical example mining method that utilizes a data-driven approach to estimate the rareness of the trajectories. By combining the rareness estimation of observations with whole trajectories, the proposed method effectively identifies a subset of data that is relatively hard to predict BEFORE feeding them to a specific prediction model. The experimental results show that the mined subset has higher prediction error when applied to different downstream prediction models, which reaches +108.1% error (greater than two times compared to the average on dataset) when mining 5% samples. Further analysis indicates that the mined critical examples include uncommon cases such as sudden brake and cancelled lane-change, which helps to better understand and improve the performance of prediction models.

Figures (6)

• Figure 1: Long-tail phenomenon in different trajectory prediction models. Nearly 20% of samples hold extremely higher errors than average. The CS-LSTM and V-LSTM are taken from deo2018convolutional, while RA-GAT is introduced in ding2021ra.
• Figure 2: Outline of the proposed critical example mining method. The details of ① Feature Extraction will be introduced in Sec.\ref{['Subsec_feature']}), the ② Probability Density Estimation will be expanded in Sec.\ref{['Subsec_flow']}.
• Figure 3: Relevant scene vehicles for TV's feature extraction.
• Figure 4: Learning results of $\boldsymbol{\mathcal{C}_{x_i}}$ , $\boldsymbol{\mathcal{C}_{z_i}}$, and $\boldsymbol{\mathcal{C}_{y_i|x_i}}$. The threshold $\delta_X$, $\delta_Z$, and $\delta_Y$ with $r=10\%$ are showed by the red dashed-lines.
• Figure 5: (a) Prediction error coverage on different models with $r=10\%$. (b) Ablation results on different $\lambda$ settings.