DriveCache: Action-Aware Caching for Driving World Model Inference
2026-08-17 • Artificial Intelligence
Artificial IntelligenceComputer Vision and Pattern Recognition
AI summaryⓘ
The authors focus on improving video generation models that predict how driving scenes will look in the near future, which helps in designing self-driving cars. They point out that current diffusion-based models are slow because they repeatedly perform complex calculations. Their method, called DriveCache, uses information about planned driving actions to speed up this process without retraining the model. It smartly decides when to reuse calculations and when to refresh them to keep results accurate. Tests show that DriveCache balances speed and video quality better than existing caching methods.
driving video generationdiffusion modelscache accelerationego motionplanned trajectorydynamic programmingdenoising stepsautonomous drivingfidelity-efficiency trade-offcausal drift
Authors
Jianchun Yang, Jian Liang, Xianda Guo, Pinhan Fu, Yanlun Peng, Conglang Zhang, Wenke Huang, Mang Ye
Abstract
Driving video generation models support autonomous-driving development by predicting controllable future scenes for simulation, planning evaluation, and offline data generation. Diffusion-based driving generators repeatedly evaluate large backbones across denoising steps, which limits generation throughput. Existing diffusion acceleration methods reduce this cost, but general-purpose designs omit driving signals available before generation, such as ego speed and planned trajectories. Experiments across driving motions show that cache tolerance varies with ego translation and rotation, denoising progress, and consecutive reuse length. We propose DriveCache, a training-free, action-aware controller that uses planned motion to allocate reuse across scenes and dynamic programming to place it across denoising steps under a calibrated response budget. A causal drift check refreshes features and replans the remaining schedule when generation departs from calibration. Across three generator configurations, DriveCache improves the overall fidelity-efficiency trade-off over evaluated cache methods. Our code will be publicly available.