Efficient Motion Prediction: A Lightweight & Accurate Trajectory Prediction Model With Fast Training and Inference Speed

TL;DR

A new efficient motion prediction model is proposed, which achieves highly competitive benchmark results while training only a few hours on a single GPU, and can easily be applied to custom datasets.

Abstract

For efficient and safe autonomous driving, it is essential that autonomous vehicles can predict the motion of other traffic agents. While highly accurate, current motion prediction models often impose significant challenges in terms of training resource requirements and deployment on embedded hardware. We propose a new efficient motion prediction model, which achieves highly competitive benchmark results while training only a few hours on a single GPU. Due to our lightweight architectural choices and the focus on reducing the required training resources, our model can easily be applied to custom datasets. Furthermore, its low inference latency makes it particularly suitable for deployment in autonomous applications with limited computing resources.

Figures (4)

• Figure 1: Resource and accuracy trade-off of our model compared to Forecast-MAE cheng2023forecast and MacFormer feng2023macformer. Top: Required training time on a single NVIDIA V100 GPU vs. the minFDE$_6$ on the Argoverse 2 validation set. Bottom: Inference latency for predicting 32 scenarios on a single NVIDIA RTX 2080 TI vs. the brier-minFDE$_6$ on the AV2 test set.
• Figure 2: General overview of our EMP model architecture. $||$ denotes concatenation.
• Figure 3: Decode module architectures: simple MLP-based decoder (for EMP-M, top) and DETR-like decoder (for EMP-D, bottom).
• Figure 4: Exemplary EMP-D results on AV2 for 2 vehicles (left, middle) and 1 pedestrian (right), showing the focal agent (orange), predictions (orange), ground truth (turquoise), history (blue), other agents (brown) and lane centerlines (black).