SAPI: Surroundings-Aware Vehicle Trajectory Prediction at Intersections

TL;DR

The paper addresses intersection trajectory prediction under dynamic surroundings. It proposes SAPI, which uses an abstract two-channel environment representation derived from real-time maps, right-of-way, and surrounding traffic, processed by a scene encoder, a sequence encoder, a refiner, and a GRU-based decoder. On a real-world proprietary Arizona dataset, SAPI achieves a 6-second horizon with $ADE$ of $1.84$ m and $FDE$ of $4.32$ m, outperforming baselines while maintaining a lightweight footprint. The approach offers robustness across scenarios and can be extended to incorporate additional cues such as traffic-light status, enhancing practical deployment for autonomous vehicle planning.

Abstract

In this work we propose a deep learning model, i.e., SAPI, to predict vehicle trajectories at intersections. SAPI uses an abstract way to represent and encode surrounding environment by utilizing information from real-time map, right-of-way, and surrounding traffic. The proposed model consists of two convolutional network (CNN) and recurrent neural network (RNN)-based encoders and one decoder. A refiner is proposed to conduct a look-back operation inside the model, in order to make full use of raw history trajectory information. We evaluate SAPI on a proprietary dataset collected in real-world intersections through autonomous vehicles. It is demonstrated that SAPI shows promising performance when predicting vehicle trajectories at intersection, and outperforms benchmark methods. The average displacement error(ADE) and final displacement error(FDE) for 6-second prediction are 1.84m and 4.32m respectively. We also show that the proposed model can accurately predict vehicle trajectories in different scenarios.

Figures (4)

• Figure 1: Environment representation at a time step
• Figure 2: Proposed SAPI architecture
• Figure 3: Displacement error at each prediction time step
• Figure 4: Model result illustration on different scenarios.