MGTR: Multi-Granular Transformer for Motion Prediction with LiDAR
This addresses motion prediction for autonomous vehicles, but it appears incremental as it builds on existing transformer and LiDAR methods.
The paper tackles motion prediction for autonomous driving by proposing a Multi-Granular Transformer (MGTR) framework that uses LiDAR data, achieving state-of-the-art performance by ranking 1st on the Waymo Open Dataset benchmark.
Motion prediction has been an essential component of autonomous driving systems since it handles highly uncertain and complex scenarios involving moving agents of different types. In this paper, we propose a Multi-Granular TRansformer (MGTR) framework, an encoder-decoder network that exploits context features in different granularities for different kinds of traffic agents. To further enhance MGTR's capabilities, we leverage LiDAR point cloud data by incorporating LiDAR semantic features from an off-the-shelf LiDAR feature extractor. We evaluate MGTR on Waymo Open Dataset motion prediction benchmark and show that the proposed method achieved state-of-the-art performance, ranking 1st on its leaderboard (https://waymo.com/open/challenges/2023/motion-prediction/).