# Rethinking the Spatio-Temporal Alignment of End-to-End 3D Perception

arXiv: https://arxiv.org/abs/2512.23635 · HTML: https://mlsanity.com/paper/2512.23635/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Builds on 15 of its 26 references that are in the map.
- First posted on arXiv 29 Dec 2025.

## Builds on (most cited first, 15 of 15)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [Deformable DETR: Deformable Transformers for End-to-End Object Detection](https://mlsanity.com/paper/2010.04159/) (2020)
- [DETR3D: 3D Object Detection from Multi-view Images via 3D-to-2D Queries](https://mlsanity.com/paper/2110.06922/) (2021)
- [Motion Transformer with Global Intention Localization and Local Movement Refinement](https://mlsanity.com/paper/2209.13508/) (2022)
- [SparseDrive: End-to-End Autonomous Driving via Sparse Scene Representation](https://mlsanity.com/paper/2405.19620/) (2024)
- [CenterMask : Real-Time Anchor-Free Instance Segmentation](https://mlsanity.com/paper/1911.06667/) (2019)
- [SparseBEV: High-Performance Sparse 3D Object Detection from Multi-Camera Videos](https://mlsanity.com/paper/2308.09244/) (2023)
- [MUTR3D: A Multi-camera Tracking Framework via 3D-to-2D Queries](https://mlsanity.com/paper/2205.00613/) (2022)
- [AdaMixer: A Fast-Converging Query-Based Object Detector](https://mlsanity.com/paper/2203.16507/) (2022)
- [SmoothNet: A Plug-and-Play Network for Refining Human Poses in Videos](https://mlsanity.com/paper/2112.13715/) (2021)
- [CAMO-MOT: Combined Appearance-Motion Optimization for 3D Multi-Object Tracking with Camera-LiDAR Fusion](https://mlsanity.com/paper/2209.02540/) (2022)
- [ADA-Track++: End-to-End Multi-Camera 3D Multi-Object Tracking with Alternating Detection and Association](https://mlsanity.com/paper/2405.08909/) (2024)
- [MV2DFusion: Leveraging Modality-Specific Object Semantics for Multi-Modal 3D Detection](https://mlsanity.com/paper/2408.05945/) (2024)
- [S.T.A.R.-Track: Latent Motion Models for End-to-End 3D Object Tracking with Adaptive Spatio-Temporal Appearance Representations](https://mlsanity.com/paper/2306.17602/) (2023)
- [SimPB: A Single Model for 2D and 3D Object Detection from Multiple Cameras](https://mlsanity.com/paper/2403.10353/) (2024)

## Built on by (most cited first, 3 of 3)

- [Offline-Poly: A Polyhedral Framework For Offline 3D Multi-Object Tracking](https://mlsanity.com/paper/2602.13772/) (2026)
- [Fusion-Poly: A Polyhedral Framework Based on Spatial-Temporal Fusion for 3D Multi-Object Tracking](https://mlsanity.com/paper/2603.08199/) (2026)
- [MatchFusion: Explicit-Implicit Instance Matching for Spatio-Temporal Multimodal Autonomous Driving](https://mlsanity.com/paper/2609.25860/) (2026)

## Abstract (arXiv)

Spatio-temporal alignment is crucial for temporal modeling of end-to-end (E2E) perception in autonomous driving (AD), providing valuable structural and textural prior information. Existing methods typically rely on the attention mechanism to align objects across frames, simplifying the motion model with a unified explicit physical model (constant velocity, etc.). These approaches prefer semantic features for implicit alignment, challenging the importance of explicit motion modeling in the traditional perception paradigm. However, variations in motion states and object features across categories and frames render this alignment suboptimal. To address this, we propose HAT, a spatio-temporal alignment module that allows each object to adaptively decode the optimal alignment proposal from multiple hypotheses without direct supervision. Specifically, HAT first utilizes multiple explicit motion models to generate spatial anchors and motion-aware feature proposals for historical instances. It then performs multi-hypothesis decoding by incorporating semantic and motion cues embedded in cached object queries, ultimately providing the optimal alignment proposal for the target frame. On nuScenes, HAT consistently improves 3D temporal detectors and trackers across diverse baselines. It achieves state-of-the-art tracking results with 46.0% AMOTA on the test set when paired with the DETR3D detector. In an object-centric E2E AD method, HAT enhances perception accuracy (+1.3% mAP, +3.1% AMOTA) and reduces the collision rate by 32%. When semantics are corrupted (nuScenes-C), the enhancement of motion modeling by HAT enables more robust perception and planning in the E2E AD.