We introduce Diffusion Policy Policy Optimization, DPPO, an algorithmic framework including best practices for fine-tuning diffusion-based policies (e.g. Diffusion Policy) in continuous control and robot learning tasks using the policy gradient (PG) method from reinforcement learning (RL). PG methods are ubiquitous in training RL policies with other policy parameterizations; nevertheless, they had been conjectured to be less efficient for diffusion-based policies. Surprisingly, we show that DPPO achieves the strongest overall performance and efficiency for fine-tuning in common benchmarks compared to other RL methods for diffusion-based policies and also compared to PG fine-tuning of other policy parameterizations. Through experimental investigation, we find that DPPO takes advantage of unique synergies between RL fine-tuning and the diffusion parameterization, leading to structured and on-manifold exploration, stable training, and strong policy robustness. We further demonstrate the strengths of DPPO in a range of realistic settings, including simulated robotic tasks with pixel observations, and via zero-shot deployment of simulation-trained policies on robot hardware in a long-horizon, multi-stage manipulation task.

Vision-language-action (VLA) models trained on large-scale internet data and robot demonstrations have the potential to serve as generalist robot policies. However, despite their large-scale training, VLAs are often brittle to task-irrelevant visual details such as distractor objects or background colors. We introduce Bring Your Own VLA (BYOVLA): a run-time intervention scheme that (1) dynamically identifies regions of the input image that the model is sensitive to, and (2) minimally alters task-irrelevant regions to reduce the model's sensitivity using automated image editing tools. Our approach is compatible with any off the shelf VLA without model fine-tuning or access to the model's weights. Hardware experiments on language-instructed manipulation tasks demonstrate that BYOVLA enables state-of-the-art VLA models to nearly retain their nominal performance in the presence of distractor objects and backgrounds, which otherwise degrade task success rates by up to 60%. Website with additional information, videos, and code: https://aasherh.github.io/byovla/.

Vision-language-action (VLA) models provide a powerful approach to training control policies for physical systems, such as robots, by combining end-to-end learning with transfer of semantic knowledge from web-scale vision-language model (VLM) training. However, the constraints of real-time control are often at odds with the design of VLMs: the most powerful VLMs have tens or hundreds of billions of parameters, presenting an obstacle to real-time inference, and operate on discrete tokens rather than the continuous-valued outputs that are required for controlling robots. To address this challenge, recent VLA models have used specialized modules for efficient continuous control, such as action experts or continuous output heads, which typically require adding new untrained parameters to the pretrained VLM backbone. While these modules improve real-time and control capabilities, it remains an open question whether they preserve or degrade the semantic knowledge contained in the pretrained VLM, and what effect they have on the VLA training dynamics. In this paper, we study this question in the context of VLAs that include a continuous diffusion or flow matching action expert, showing that naively including such experts significantly harms both training speed and knowledge transfer. We provide an extensive analysis of various design choices, their impact on performance and knowledge transfer, and propose a technique for insulating the VLM backbone during VLA training that mitigates this issue. Videos are available at https://pi.website/research/knowledge_insulation and open-source model weights are available at https://github.com/Physical-Intelligence/openpi.

Safety is a critical component of autonomous systems and remains a challenge for learning-based policies to be utilized in the real world. In particular, policies learned using reinforcement learning often fail to generalize to novel environments due to unsafe behavior. In this paper, we propose Sim-to-Lab-to-Real to bridge the reality gap with a probabilistically guaranteed safety-aware policy distribution. To improve safety, we apply a dual policy setup where a performance policy is trained using the cumulative task reward and a backup (safety) policy is trained by solving the Safety Bellman Equation based on Hamilton-Jacobi (HJ) reachability analysis. In Sim-to-Lab transfer, we apply a supervisory control scheme to shield unsafe actions during exploration; in Lab-to-Real transfer, we leverage the Probably Approximately Correct (PAC)-Bayes framework to provide lower bounds on the expected performance and safety of policies in unseen environments. Additionally, inheriting from the HJ reachability analysis, the bound accounts for the expectation over the worst-case safety in each environment. We empirically study the proposed framework for ego-vision navigation in two types of indoor environments with varying degrees of photorealism. We also demonstrate strong generalization performance through hardware experiments in real indoor spaces with a quadrupedal robot. See https://sites.google.com/princeton.edu/sim-to-lab-to-real for supplementary material.

Large Language Models (LLMs) have demonstrated the ability to perform semantic reasoning, planning and write code for robotics tasks. However, most methods rely on pre-existing primitives (i.e. pick, open drawer) or similar examples of robot code alone, which heavily limits their scalability to new scenarios. We present PromptBook, a collection of different prompting paradigms to generate code for successfully executing new manipulation skills. We demonstrate example-based, instruction-based and chain-of-thought to write robot code; as well as a method to build the prompt leveraging LLMs and human feedback. We show PromptBook enables LLMs to write code for new low-level manipulation skills in a zero-shot manner: from picking diverse objects, opening/closing drawers, to whisking, and waving hello. We evaluate the new skills on a mobile manipulator with 83% success rate at picking, 50-71% at opening drawers and 100% at closing them. Notably, the LLM is able to infer gripper orientation for grasping a drawer handle (z-axis aligned) vs. a top-down grasp (x-axis aligned).

Unmanned aerial vehicles (UAVs) are finding use in applications that place increasing emphasis on robustness to external disturbances including extreme wind. However, traditional multirotor UAV platforms do not directly sense wind; conventional flow sensors are too slow, insensitive, or bulky for widespread integration on UAVs. Instead, drones typically observe the effects of wind indirectly through accumulated errors in position or trajectory tracking. In this work, we integrate a novel flow sensor based on micro-electro-mechanical systems (MEMS) hot-wire technology developed in our prior work [1] onto a multirotor UAV for wind estimation. Our sensor is omnidirectional (in the plane), lightweight, fast, and accurate. In order to achieve superior hover performance in windy conditions, we train a ‘wind-aware’ residual-based controller via reinforcement learning using simulated wind gusts and their aerodynamic effects on the drone. In extensive hardware experiments, we demonstrate the wind-aware controller out-performing two strong ‘wind-unaware’ baseline controllers in challenging windy conditions. See: youtu.be/KWqkH9Z-338.

We are motivated by the problem of learning policies for robotic systems with rich sensory inputs (e.g., vision) in a manner that allows us to guarantee generalization to environments unseen during training. We provide a framework for providing such generalization guarantees by leveraging a finite dataset of real-world environments in combination with a (potentially inaccurate) generative model of environments. The key idea behind our approach is to utilize the generative model in order to implicitly specify a prior over policies. This prior is updated using the real-world dataset of environments by minimizing an upper bound on the expected cost across novel environments derived via Probably Approximately Correct (PAC)-Bayes generalization theory. We demonstrate our approach on two simulated systems with nonlinear/hybrid dynamics and rich sensing modalities: (i) quadrotor navigation with an onboard vision sensor, and (ii) grasping objects using a depth sensor. Comparisons with prior work demonstrate the ability of our approach to obtain stronger generalization guarantees by utilizing generative models. We also present hardware experiments for validating our bounds for the grasping task.