… In this paper, we present Wayformer, a family of simple and homogeneous attention based
architectures for motion forecasting. Wayformer offers a compact model description consisting …

… We provide Wayformer models with the nearest 512 map polylines and RedMotion models
with the nearest 128 map polylines, and both models with a maximum of 48 surrounding …

… Reliable forecasting of the future behavior of road agents is a … discrete motion tokens and
cast multi-agent motion prediction … -agent motion prediction on the Waymo Open Motion Dataset…

… To evaluate the suitability of LiDAR to the motion forecasting task, we provide a WayFormer
baseline trained with LiDAR. Experiments show that LiDAR data brings improvement in the …

… Wayformer: Motion forecasting via simple & efficient attention networks. arXiv preprint
arXiv:2207.05844, 2022. 3 [26] Jiquan Ngiam, Benjamin Caine, Wei Han, Brandon Yang, Yuning …

… state-of-the-art methods such as Wayformer [36] with much fewer parameters. The evaluation
results of the Argoverse 2 motion forecasting benchmark are shown in Tab. II. We compare …

… We study the challenges of deploying motion forecasting models into the real world when
only … Sapp, “Wayformer: Motion forecasting via simple & efficient attention networks,” in ICRA, …

… Wayformer [46], as our motion prediction backbone and we augment it with our new design
by fusing multi-modality tokens. In the following sections, we use Wayformer as the baseline …

… Our results show that existing motion forecasting models are sensitive to deleting non-…
Across them, ST Marginal and Wayformer are the most robust models. Compared to Wayformer, …

… motion prediction model, called AMP, achieves state-of-the-art performance on the Waymo
Open Motion … – We propose an autoregressive prediction paradigm for motion forecasting. …