DriftWorld: Fast World Modeling through Drifting
RoboticsComputer Vision and Pattern RecognitionMachine Learning
Summary
The authors present DriftWorld, a new type of robot world model that can quickly imagine future actions and their results in one step, unlike slower methods that need many steps to do this. This speedup is about 17 times faster than previous diffusion-based models, making it easier for robots to plan and make decisions fast. They tested DriftWorld on various robot tasks and found it to be both accurate and efficient. Additionally, it can be used to evaluate how good robot policies are without needing real-world trials, showing strong agreement with real outcomes.
Authors
Susie Lu, Haonan Chen, Weirui Ye, Yilun Du
Abstract
Predictive world models enable robots to plan by imagining the outcomes of their actions, but their value for control hinges on generating many rollouts quickly. This creates a bottleneck for diffusion-based world models: multistep sampling makes each rollout expensive, limiting large-scale action search at inference time. We introduce DriftWorld, an action-conditioned world model based on drifting generative models. Rather than denoising iteratively at inference, DriftWorld learns an action-conditioned drift during training, allowing it to generate future frames from the current observation and a candidate action sequence in a single forward pass at 30+ fps, which is 17x faster on average than diffusion based baselines. We evaluate DriftWorld on standard vision-based robotic manipulation benchmarks, including Bridge-V2, RT-1, Language Table, Push-T, and Robomimic. By producing rollouts that are both accurate and fast, DriftWorld achieves state-of-the-art decision-making performance with far less inference time than diffusion-based world model baselines. Beyond online control, DriftWorld can also serve as an offline simulator for ranking real-world robot policies, with rollout-based scores correlating with ground truth at up to 0.99. These results show that drifting models are a strong fit for robot world modeling, where fast, high-quality imagination directly supports planning and policy evaluation.