BANet: Motion Forecasting with Boundary Aware Network
This addresses motion forecasting for autonomous vehicles by improving prediction accuracy through richer map encoding, though it is incremental as a variant of an existing method.
The paper tackles motion forecasting by proposing BANet, a Boundary-Aware Network variant of LaneGCN that encodes additional vector map elements like lane boundaries to capture traffic rule constraints, achieving first place on the Argoverse2 Motion Forecasting challenge test leaderboard.
We propose a motion forecasting model called BANet, which means Boundary-Aware Network, and it is a variant of LaneGCN. We believe that it is not enough to use only the lane centerline as input to obtain the embedding features of the vector map nodes. The lane centerline can only provide the topology of the lanes, and other elements of the vector map also contain rich information. For example, the lane boundary can provide traffic rule constraint information such as whether it is possible to change lanes which is very important. Therefore, we achieved better performance by encoding more vector map elements in the motion forecasting model.We report our results on the 2022 Argoverse2 Motion Forecasting challenge and rank 1st on the test leaderboard.