# Drift-Based Policy Optimization: Native One-Step Policy Learning for Online Robot Control

arXiv: https://arxiv.org/abs/2604.03540 · HTML: https://mlsanity.com/paper/2604.03540/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- Placed in Diffusion and flow policies for robotics, one of 14 directions in Diffusion models.
- Builds on 36 of its 37 references that are in the map.
- First posted on arXiv 4 Apr 2026.

## Placed in

- Diffusion models / [Diffusion and flow policies for robotics](https://mlsanity.com/diffusion/diffusion-and-flow-policies-for-robotics/)
- Deep reinforcement learning methods / [RL with diffusion and flow models](https://mlsanity.com/deeprl/rl-with-diffusion-and-flow-models/)

## Builds on (most cited first, 30 of 36)

- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [Flow Matching for Generative Modeling](https://mlsanity.com/paper/2210.02747/) (2022)
- [Scaling Rectified Flow Transformers for High-Resolution Image Synthesis](https://mlsanity.com/paper/2403.03206/) (2024)
- [Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow](https://mlsanity.com/paper/2209.03003/) (2022)
- [Diffusion Policy: Visuomotor Policy Learning via Action Diffusion](https://mlsanity.com/paper/2303.04137/) (2023)
- [Consistency Models](https://mlsanity.com/paper/2303.01469/) (2023)
- [Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware](https://mlsanity.com/paper/2304.13705/) (2023)
- [D4RL: Datasets for Deep Data-Driven Reinforcement Learning](https://mlsanity.com/paper/2004.07219/) (2020)
- [SiT: Exploring Flow and Diffusion-based Generative Models with Scalable Interpolant Transformers](https://mlsanity.com/paper/2401.08740/) (2024)
- [What Matters in Learning from Offline Human Demonstrations for Robot Manipulation](https://mlsanity.com/paper/2108.03298/) (2021)
- [Learning Complex Dexterous Manipulation with Deep Reinforcement Learning and Demonstrations](https://mlsanity.com/paper/1709.10087/) (2017)
- [Flow-GRPO: Training Flow Matching Models via Online RL](https://mlsanity.com/paper/2505.05470/) (2025)
- [Mean Flows for One-step Generative Modeling](https://mlsanity.com/paper/2505.13447/) (2025)
- [Improved Techniques for Training Consistency Models](https://mlsanity.com/paper/2310.14189/) (2023)
- [One Step Diffusion via Shortcut Models](https://mlsanity.com/paper/2410.12557/) (2024)
- [Diffusion Policy Policy Optimization](https://mlsanity.com/paper/2409.00588/) (2024)
- [VLA-RL: Towards Masterful and General Robotic Manipulation with Scalable Reinforcement Learning](https://mlsanity.com/paper/2505.18719/) (2025)
- [Consistency Policy: Accelerated Visuomotor Policies via Consistency Distillation](https://mlsanity.com/paper/2405.07503/) (2024)
- [Generative Modeling via Drifting](https://mlsanity.com/paper/2602.04770/) (2026)
- [Improved Mean Flows: On the Challenges of Fastforward Generative Models](https://mlsanity.com/paper/2512.02012/) (2025)
- [FlowPolicy: Enabling Fast and Robust 3D Flow-based Policy via Consistency Flow Matching for Robot Manipulation](https://mlsanity.com/paper/2412.04987/) (2024)
- [Flow Matching Policy Gradients](https://mlsanity.com/paper/2507.21053/) (2025)
- [ReinFlow: Fine-tuning Flow Matching Policy with Online Reinforcement Learning](https://mlsanity.com/paper/2505.22094/) (2025)
- [One-Step Diffusion Policy: Fast Visuomotor Policies via Diffusion Distillation](https://mlsanity.com/paper/2410.21257/) (2024)
- [Diffuse and Disperse: Image Generation with Representation Regularization](https://mlsanity.com/paper/2506.09027/) (2025)
- [AdaFlow: Imitation Learning with Variance-Adaptive Flow-Based Policies](https://mlsanity.com/paper/2402.04292/) (2024)
- [ManiFlow: A General Robot Manipulation Policy via Consistency Flow Training](https://mlsanity.com/paper/2509.01819/) (2025)
- [Meta-World+: An Improved, Standardized, RL Benchmark](https://mlsanity.com/paper/2505.11289/) (2025)
- [MP1: MeanFlow Tames Policy Learning in 1-step for Robotic Manipulation](https://mlsanity.com/paper/2507.10543/) (2025)
- [SplitMeanFlow: Interval Splitting Consistency in Few-Step Generative Modeling](https://mlsanity.com/paper/2507.16884/) (2025)

## Built on by (most cited first, 5 of 5)

- [One-Step Generative Modeling via Wasserstein Gradient Flows](https://mlsanity.com/paper/2605.11755/) (2026)
- [Drifting Field Policy: A One-Step Generative Policy via Wasserstein Gradient Flow](https://mlsanity.com/paper/2605.07727/) (2026)
- [CAC-VLA: Context-Gated Action Conditioning for Vision-Language-Action Models](https://mlsanity.com/paper/2607.04816/) (2026)
- [DriftingVLA: Native One-Step Vision-Language-Action Generation via Per-Dimension Temporal Drifting](https://mlsanity.com/paper/2608.29749/) (2026)
- [Occupancy Modeling via Drifting: One-Step Generation Enables Direct Policy Extraction](https://mlsanity.com/paper/2610.19883/) (2026)

## Abstract (arXiv)

Diffusion policies effectively model multimodal action distributions for robotic manipulation, but their iterative denoising requires tens to hundreds of network function evaluations (NFEs) for each control prediction, limiting their applicability to high-frequency closed-loop control and online reinforcement learning (RL). We present a two-stage framework for native one-step generative policies that transfers iterative refinement from inference to training. First, Drift-Based Policy (DBP) uses a fixed-point drifting objective to internalize corrective dynamics into the model parameters, producing multimodal action chunks with a single network evaluation by design. Second, Drift-Based Policy Optimization (DBPO) augments the pretrained backbone with a stochastic interface that provides exact conditional rollout likelihoods for PPO-style on-policy updates while preserving 1-NFE deployment. On the 12-task Diffusion Policy suite, DBP improves the average success rate from 0.79 to 0.83 while reducing inference from 100 NFEs to 1. Across 37 point-cloud manipulation tasks, DBP achieves an average success rate of 88.4%, surpassing the leading 1-NFE baseline OMP at 82.3%. DBPO further improves pretrained one-step policies through stable online fine-tuning on RoboMimic and D4RL. On a physical dual-arm UR5 platform, DBP achieves 123/150 successes (82%) with an average end-to-end latency of 9.5 ms, compared with MP1's 89/150 successes (59%) under the same setup. Code is available at https://github.com/YuxuanGao0822/DBPO.