Driving · 2026-11-06

FreeOcc: Decoupling Ego-Motion for Efficient 4D Occupancy Forecasting via Continuous Flow Matching

Zeping Zhang, Zhuoya Zhao, Samy Metari, Robert Lagani\`ere

arXiv:2611.08128PDF

AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.

TL;DR

FreeOcc is a lightweight continuous-time 4D occupancy forecasting model that explicitly decouples ego-motion using SE(2) geometric warping and then predicts residual dynamics with physics-constrained conditional flow matching. It uses uncertainty-aware blending to route well-observed voxels to a fast decoder and ambiguous regions to a continuous ODE solver. The paper reports 27 FPS for FreeOcc-E with 28.3M parameters.

Why it matters

The authors argue that many 4D occupancy world models spend capacity on learned temporal alignment/ego-motion disentanglement, even though future ego motion is available from a motion planner. FreeOcc replaces attention-heavy alignment with explicit SE(2) warping, enabling attention-free temporal fusion, uncertainty-aware compute routing, and physics constraints applied directly to residual agent kinematics.

Method

  • Explicit ego-motion decoupling: use SE(2) ego-aware warping to map the fused BEV context to future coordinate frames, isolating residual scene changes.
  • Physics-constrained continuous flow matching: train a 3D U-Net velocity network with flow-matching, conditioned on the decoupled residual BEV prior, with regularization (e.g., velocity smoothness, mass conservation, ray-traced free-space).
  • Uncertainty-aware ensemble blending: a discriminative decoder outputs per-voxel log-variance; confident voxels use a single-pass decoder while highly uncertain regions are handled by the conditional flow-matching ODE.

Limitation

Concurrent works SuperOcc and DTT are not directly applicable under the authors’ protocol: SuperOcc targets current-frame 3D prediction (not 4D forecasting) and DTT lacks publicly released code.

Abstract (from arXiv)

Current 4D occupancy world models use heavy cross-attention or deep recurrence to disentangle ego-motion from scene dynamics. This requires significant computational capacity. However, this effort is largely redundant since the ego-vehicle's intended future motion is provided by the downstream motion planner. We present FreeOcc, a lightweight continuous-time world model that introduces explicit $SE(2)$ geometric decoupling into a conditional flow-matching framework. By mathematically factoring out self-motion, we create a purely dynamic feature space. This space supports attention-free temporal fusion and allows physics-constrained ODEs to operate strictly on residual agent kinematics. To maximize deployment efficiency, FreeOcc introduces an uncertainty-aware blending mechanism. This routes well-observed voxels to a single-pass decoder, reserving the ODE solver exclusively for ambiguous, high-uncertainty regions. Requiring only 28.3M parameters-roughly 40% of comparable models-our optimal ensemble mode (FreeOcc-E) runs at 27 FPS. On nuScenes, FreeOcc achieves state-of-the-art forecasting performance with a 0.47 m Near Field Chamfer Distance (NFCD) and 32.1% mIoU. Furthermore, the model transfers zero-shot to SemanticKITTI. It also proves highly robust, retaining 92.8% of its ideal mIoU under significant sensor packet loss and timestamp jitter.

Related papers

Publication record

Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.

First public release: 2026-11-05 · Indexed by Pub-AI: 2026-11-06 UTC

preprint

FreeOcc: Decoupling Ego-Motion for Efficient 4D Occupancy Forecasting via Continuous Flow Matching

Zeping Zhang, Zhuoya Zhao, Samy Metari, Robert Laganière

Preprint posted: 2026-11-05 · Original record ↗

arxiv
2611.08128
Explicit source versions (1)
  • v1 · 2026-11-05
  1. Preprint available

    Source date: 2026-11-05 · First detected: 2026-11-06 UTC · Historical baseline

    Source evidence (3)
Abstract · arxiv

Current 4D occupancy world models use heavy cross-attention or deep recurrence to disentangle ego-motion from scene dynamics. This requires significant computational capacity. However, this effort is largely redundant since the ego-vehicle's intended future motion is provided by the downstream motion planner. We present FreeOcc, a lightweight continuous-time world model that introduces explicit $SE(2)$ geometric decoupling into a conditional flow-matching framework. By mathematically factoring out self-motion, we create a purely dynamic feature space. This space supports attention-free temporal fusion and allows physics-constrained ODEs to operate strictly on residual agent kinematics. To maximize deployment efficiency, FreeOcc introduces an uncertainty-aware blending mechanism. This routes well-observed voxels to a single-pass decoder, reserving the ODE solver exclusively for ambiguous, high-uncertainty regions. Requiring only 28.3M parameters-roughly 40% of comparable models-our optimal ensemble mode (FreeOcc-E) runs at 27 FPS. On nuScenes, FreeOcc achieves state-of-the-art forecasting performance with a 0.47 m Near Field Chamfer Distance (NFCD) and 32.1% mIoU. Furthermore, the model transfers zero-shot to SemanticKITTI. It also proves highly robust, retaining 92.8% of its ideal mIoU under significant sensor packet loss and timestamp jitter.