Map-Free Trajectory Prediction with Map Distillation and Hierarchical Encoding

Xiaodong Liu,Yucheng Xing,Xin Wang

2024-11-17

Abstract:Reliable motion forecasting of surrounding agents is essential for ensuring the safe operation of autonomous vehicles. Many existing trajectory prediction methods rely heavily on high-definition (HD) maps as strong driving priors. However, the availability and accuracy of these priors are not guaranteed due to substantial costs to build, localization errors of vehicles, or ongoing road constructions. In this paper, we introduce MFTP, a Map-Free Trajectory Prediction method that offers several advantages. First, it eliminates the need for HD maps during inference while still benefiting from map priors during training via knowledge distillation. Second, we present a novel hierarchical encoder that effectively extracts spatial-temporal agent features and aggregates them into multiple trajectory queries. Additionally, we introduce an iterative decoder that sequentially decodes trajectory queries to generate the final predictions. Extensive experiments show that our approach achieves state-of-the-art performance on the Argoverse dataset under the map-free setting.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to perform reliable trajectory prediction in an autonomous driving system without relying on high - definition maps (HD maps). Specifically, most of the existing trajectory prediction methods rely on high - definition maps as strong prior knowledge, but the availability and accuracy of these maps cannot be guaranteed in practical applications, which limits their application in real - driving scenarios. Therefore, this research proposes a method named MFTP (Map - Free Trajectory Prediction), aiming to eliminate the need for high - definition maps during the inference process, and at the same time utilize map prior knowledge in the training stage through knowledge distillation technology, so as to achieve map - free trajectory prediction while maintaining high performance.

The main contributions of the paper include:
1. Proposing a method for trajectory prediction without the need for high - definition maps. This method utilizes map prior knowledge through knowledge distillation technology in the training stage, and does not rely on maps at all in the inference stage.
2. Introducing a novel hierarchical encoder, which can extract multi - level spatio - temporal features of agents (such as vehicles, pedestrians, etc.) and aggregate them into multiple trajectory queries.
3. Designing an iterative decoder, which can sequentially decode trajectory queries to generate the final prediction results.
4. Experiments show that this method achieves new state - of - the - art performance in the map - free setting on the Argoverse dataset.

Through these innovations, MFTP not only improves the performance of map - free trajectory prediction, but also emphasizes the importance of reducing the dependence on external data sources (such as high - definition maps) in actual autonomous driving systems.