MGTR: LiDARを用いたマルチグラニュラートランスフォーマによる運動予測
MGTR: Multi-Granular Transformer for Motion Prediction with LiDAR
異なる種類の交通エージェントに対し、異なる粒度の文脈特徴を活用するエンコーダ・デコーダ型トランスフォーマを提案し、LiDARの意味特徴を組み込むことでWaymo Open Datasetの運動予測ベンチマークで最高性能を達成した。
著者: Yiqian Gan, Hao Xiao, Yizhe Zhao, Ethan Zhang, Zhe Huang, Xin Ye, Lingting Ge
分類: cs.CV, cs.RO
原文アブストラクト
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/).
