Fast and accurate 3D scene prediction improves autonomous vehicle response
Fast-Slow Evolutionary Occupancy Prediction via Controlled Dynamics
Computer Vision and Pattern Recognition
Summary
Self-driving cars need to quickly and accurately understand their surroundings to navigate safely. Existing methods often trade off speed for detail or vice versa. The authors propose a new way to combine fast but less detailed models with slower, more accurate ones to get the best of both worlds. They treat the prediction process as a system that evolves over time, adapting flexibly to different speeds and changes. Their approach consistently improves accuracy while being faster than slower models alone.
What this means in practice
- •For autonomous vehicle engineers: Enable faster and more accurate 3D scene understanding for safer and more responsive vehicle navigation.
- •For robotics system integrators: Allow robots to combine fast and slow sensing data to better model environmental changes in real time.
Authors
Runhe Yang, Zhiyuan Zhou, Yuxiang Yan, Xuetong Yang, Yikai Pan, Jiacheng Tang, Zhuolin He, Jianhua Han, Hang Xu, Jian Pu
Abstract
3D semantic occupancy prediction provides dense geometric and semantic scene representations for autonomous driving, where both high accuracy and low response latency are crucial for safe downstream forecasting and planning. Existing methods usually exhibit different accuracy-latency characteristics. Performance-oriented models provide stronger geometric and semantic predictions, while deployment-friendly models provide faster responses. This motivates EvoOcc, a fast-slow evolutionary occupancy prediction framework that jointly exploits their complementary strengths to improve the accuracy-latency balance without changing the upstream architectures. EvoOcc models evolutionary occupancy prediction as a continuous time controlled dynamical system, enabling flexible adaptation to variable slow system delays and irregular fast system arrivals. The framework further decomposes the evolution process into dual state evolution where ego state evolution compensates for the deterministic coordinate shifts and scene state evolution models the genuine geometric and semantic scene changes. Extensive experiments on three different fast-slow system pairs and irregular prediction arrivals show that EvoOcc consistently improves over fast system baselines while maintaining much lower latency than the slow system, indicating a favorable accuracy-latency balance and stable behavior under practical inference situations.