CascadeOcc: Rethinking 3D Occupancy World Models with Cascaded VQ Representations

2026-06-25

AI Summary

This work addresses the overreliance of existing 3D occupancy world models on external modalities or large language models, which often overlooks the inherent structural potential of occupancy representations themselves. The authors propose a novel autoregressive 3D occupancy world model that leverages cascaded vector quantization (Cascaded VQ) and a multi-scale spatial architecture to generate scene details progressively from coarse to fine. To capture complex spatiotemporal dynamics, they introduce TimeMixer, a mechanism that explicitly models dual-level dependencies across both space and time. This approach uniquely integrates cascaded VQ with hierarchical spatiotemporal modeling, significantly enhancing the intrinsic structural expressiveness of occupancy representations without relying on external signals. Evaluated on 4D occupancy prediction and motion planning benchmarks, the method outperforms current vision-centric approaches, demonstrating the effectiveness of optimizing internal representation capacity.

Abstract

This letter proposes CascadeOcc, a novel occupancy world model that prioritizes intrinsic structural hierarchy over extrinsic auxiliary modalities for autonomous driving. Occupancy world models -- forecasting the future driving environment and planning the driving trajectory -- effectively bridge perception and planning, but current approaches often heavily rely on external modalities or large language models, failing to fully exploit the inherent structural potential of occupancy representations themselves. To enhance representational capacity for complex 3D scenes, we integrate a cascaded Vector Quantized (VQ) mechanism into an autoregressive framework. Following a coarse-to-fine principle, CascadeOcc progressively refines fine-grained details from global structures through a multi-scale architecture. Additionally, we incorporate a TimeMixer to capture multi-scale temporal dependencies, establishing a dual-hierarchy mechanism in both space and time. Experimental results on 4D occupancy forecasting and motion planning benchmarks demonstrate that CascadeOcc achieves superior performance among vision-centric approaches, validating that optimizing inherent representations is a powerful alternative to relying on external foundation models.

Problem

• 3D occupancy
• world models
• structural hierarchy
• autonomous driving
• representation learning

Innovation

• Cascaded VQ
• 3D Occupancy World Model
• Coarse-to-Fine Representation
• TimeMixer
• Autoregressive Framework