CascadeOcc: Rethinking 3D Occupancy World Models With Cascaded VQ Representations
- 1 January 2026
- journal article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Signal Processing Letters
- Vol. 33 (10709908) , 1506-1510
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.Keywords
All Related Versions
Funding Information
- Korea Research Institute for defense Technology planning
- Defense Innovation Vanguard Enterprise Project
- Defense Acquisition Program Administration (R230206)
- Technology Innovation Program (RS-2024-00445759)
- Development of Navigation Technology Utilizing Visual Information
- Ministry of Trade
- LG AI STAR Talent Development Program for Leading Large-Scale Generative AI Models (RS-2025-25442149)
This publication has 24 references indexed in Scilit:
- FSF-Net: Enhance 4D occupancy forecasting with coarse BEV scene flow for autonomous drivingPattern Recognition, 2025
- OccWorld: Learning a 3D Occupancy World Model for Autonomous Driving2024
- Tri-Perspective View for Vision-Based 3D Semantic Occupancy Prediction2023
- VoxFormer: Sparse Voxel Transformer for Camera-Based 3D Semantic Scene Completion2023
- Autoregressive Image Generation using Residual Quantization2022
- MonoScene: Monocular 3D Semantic Scene Completion2022
- Lift, Splat, Shoot: Encoding Images from Arbitrary Camera Rigs by Implicitly Unprojecting to 3D2020
- Cascade Cost Volume for High-Resolution Multi-View Stereo and Stereo Matching2020
- nuScenes: A Multimodal Dataset for Autonomous Driving2020
- Feature Pyramid Networks for Object Detection2017