Saved in:
Bibliographic Details
Main Authors: Du, Jiayuan, Zhao, Yiming, Guo, Zhenglong, Pan, Yong, Hou, Wenbo, Hao, Zhihui, Zhan, Kun, Chen, Qijun
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2511.22039
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910127039184896
author Du, Jiayuan
Zhao, Yiming
Guo, Zhenglong
Pan, Yong
Hou, Wenbo
Hao, Zhihui
Zhan, Kun
Chen, Qijun
author_facet Du, Jiayuan
Zhao, Yiming
Guo, Zhenglong
Pan, Yong
Hou, Wenbo
Hao, Zhihui
Zhan, Kun
Chen, Qijun
contents This paper introduces a novel architecture for trajectory-conditioned forecasting of future 3D scene occupancy. In contrast to methods that rely on variational autoencoders (VAEs) to generate discrete occupancy tokens, which inherently limit representational capacity, our approach predicts multi-frame future occupancy in an end-to-end manner directly from raw image features. Inspired by the success of attention-based transformer architectures in foundational vision and language models such as GPT and VGGT, we employ a sparse occupancy representation that bypasses the intermediate bird's eye view (BEV) projection and its explicit geometric priors. This design allows the transformer to capture spatiotemporal dependencies more effectively. By avoiding both the finite-capacity constraint of discrete tokenization and the structural limitations of BEV representations, our method achieves state-of-the-art performance on the nuScenes benchmark for 1-3 second occupancy forecasting, outperforming existing approaches by a significant margin. Furthermore, it demonstrates robust scene dynamics understanding, consistently delivering high accuracy under arbitrary future trajectory conditioning.
format Preprint
id arxiv_https___arxiv_org_abs_2511_22039
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle SparseWorld-TC: Trajectory-Conditioned Sparse Occupancy World Model
Du, Jiayuan
Zhao, Yiming
Guo, Zhenglong
Pan, Yong
Hou, Wenbo
Hao, Zhihui
Zhan, Kun
Chen, Qijun
Computer Vision and Pattern Recognition
This paper introduces a novel architecture for trajectory-conditioned forecasting of future 3D scene occupancy. In contrast to methods that rely on variational autoencoders (VAEs) to generate discrete occupancy tokens, which inherently limit representational capacity, our approach predicts multi-frame future occupancy in an end-to-end manner directly from raw image features. Inspired by the success of attention-based transformer architectures in foundational vision and language models such as GPT and VGGT, we employ a sparse occupancy representation that bypasses the intermediate bird's eye view (BEV) projection and its explicit geometric priors. This design allows the transformer to capture spatiotemporal dependencies more effectively. By avoiding both the finite-capacity constraint of discrete tokenization and the structural limitations of BEV representations, our method achieves state-of-the-art performance on the nuScenes benchmark for 1-3 second occupancy forecasting, outperforming existing approaches by a significant margin. Furthermore, it demonstrates robust scene dynamics understanding, consistently delivering high accuracy under arbitrary future trajectory conditioning.
title SparseWorld-TC: Trajectory-Conditioned Sparse Occupancy World Model
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2511.22039