Animesh Garg

Mayank Mittal, Calvin Yu, Qinxi Yu et al. · cmu, eth-zurich

We present Orbit, a unified and modular framework for robot learning powered by NVIDIA Isaac Sim. It offers a modular design to easily and efficiently create robotic environments with photo-realistic scenes and high-fidelity rigid and deformable body simulation. With Orbit, we provide a suite of benchmark tasks of varying difficulty -- from single-stage cabinet opening and cloth folding to multi-stage tasks such as room reorganization. To support working with diverse observations and action spaces, we include fixed-arm and mobile manipulators with different physically-based sensors and motion generators. Orbit allows training reinforcement learning policies and collecting large demonstration datasets from hand-crafted or expert solutions in a matter of minutes by leveraging GPU-based parallelization. In summary, we offer an open-sourced framework that readily comes with 16 robotic platforms, 4 sensor modalities, 10 motion generators, more than 20 benchmark tasks, and wrappers to 4 learning libraries. With this framework, we aim to support various research areas, including representation learning, reinforcement learning, imitation learning, and task and motion planning. We hope it helps establish interdisciplinary collaborations in these communities, and its modularity makes it easily extensible for more tasks and applications in the future.

Mayank Mittal, Pascal Roth, James Tigue et al. · nvidia

We present Isaac Lab, the natural successor to Isaac Gym, which extends the paradigm of GPU-native robotics simulation into the era of large-scale multi-modal learning. Isaac Lab combines high-fidelity GPU parallel physics, photorealistic rendering, and a modular, composable architecture for designing environments and training robot policies. Beyond physics and rendering, the framework integrates actuator models, multi-frequency sensor simulation, data collection pipelines, and domain randomization tools, unifying best practices for reinforcement and imitation learning at scale within a single extensible platform. We highlight its application to a diverse set of challenges, including whole-body control, cross-embodiment mobility, contact-rich and dexterous manipulation, and the integration of human demonstrations for skill acquisition. Finally, we discuss upcoming integration with the differentiable, GPU-accelerated Newton physics engine, which promises new opportunities for scalable, data-efficient, and gradient-based approaches to robot learning. We believe Isaac Lab's combination of advanced simulation capabilities, rich sensing, and data-center scale execution will help unlock the next generation of breakthroughs in robotics research.

Ishika Singh, Valts Blukis, Arsalan Mousavian et al. · gatech, nvidia

Task planning can require defining myriad domain knowledge about the world in which a robot needs to act. To ameliorate that effort, large language models (LLMs) can be used to score potential next actions during task planning, and even generate action sequences directly, given an instruction in natural language with no additional domain information. However, such methods either require enumerating all possible next steps for scoring, or generate free-form text that may contain actions not possible on a given robot in its current context. We present a programmatic LLM prompt structure that enables plan generation functional across situated environments, robot capabilities, and tasks. Our key insight is to prompt the LLM with program-like specifications of the available actions and objects in an environment, as well as with example programs that can be executed. We make concrete recommendations about prompt structure and generation constraints through ablation experiments, demonstrate state of the art success rates in VirtualHome household tasks, and deploy our method on a physical robot arm for tabletop tasks. Website at progprompt.github.io

Amir Rasouli, Randy Goebel, Matthew E. Taylor et al. · gatech, nvidia

Driving SMARTS is a regular competition designed to tackle problems caused by the distribution shift in dynamic interaction contexts that are prevalent in real-world autonomous driving (AD). The proposed competition supports methodologically diverse solutions, such as reinforcement learning (RL) and offline learning methods, trained on a combination of naturalistic AD data and open-source simulation platform SMARTS. The two-track structure allows focusing on different aspects of the distribution shift. Track 1 is open to any method and will give ML researchers with different backgrounds an opportunity to solve a real-world autonomous driving challenge. Track 2 is designed for strictly offline learning methods. Therefore, direct comparisons can be made between different methods with the aim to identify new promising research directions. The proposed setup consists of 1) realistic traffic generated using real-world data and micro simulators to ensure fidelity of the scenarios, 2) framework accommodating diverse methods for solving the problem, and 3) baseline method. As such it provides a unique opportunity for the principled investigation into various aspects of autonomous vehicle deployment.

Jie Xu, Viktor Makoviychuk, Yashraj Narang et al. · gatech, nvidia

Deep reinforcement learning can generate complex control policies, but requires large amounts of training data to work effectively. Recent work has attempted to address this issue by leveraging differentiable simulators. However, inherent problems such as local minima and exploding/vanishing numerical gradients prevent these methods from being generally applied to control tasks with complex contact-rich dynamics, such as humanoid locomotion in classical RL benchmarks. In this work we present a high-performance differentiable simulator and a new policy learning algorithm (SHAC) that can effectively leverage simulation gradients, even in the presence of non-smoothness. Our learning algorithm alleviates problems with local minima through a smooth critic function, avoids vanishing/exploding gradients through a truncated learning window, and allows many physical environments to be run in parallel. We evaluate our method on classical RL control tasks, and show substantial improvements in sample efficiency and wall-clock time over state-of-the-art RL and differentiable simulation-based algorithms. In addition, we demonstrate the scalability of our method by applying it to the challenging high-dimensional problem of muscle-actuated locomotion with a large action space, achieving a greater than 17x reduction in training time over the best-performing established RL algorithm.

Satya Krishna Gorti, Noel Vouitsis, Junwei Ma et al. · gatech, nvidia

In text-video retrieval, the objective is to learn a cross-modal similarity function between a text and a video that ranks relevant text-video pairs higher than irrelevant pairs. However, videos inherently express a much wider gamut of information than texts. Instead, texts often capture sub-regions of entire videos and are most semantically similar to certain frames within videos. Therefore, for a given text, a retrieval model should focus on the text's most semantically similar video sub-regions to make a more relevant comparison. Yet, most existing works aggregate entire videos without directly considering text. Common text-agnostic aggregations schemes include mean-pooling or self-attention over the frames, but these are likely to encode misleading visual information not described in the given text. To address this, we propose a cross-modal attention model called X-Pool that reasons between a text and the frames of a video. Our core mechanism is a scaled dot product attention for a text to attend to its most semantically similar frames. We then generate an aggregated video representation conditioned on the text's attention weights over the frames. We evaluate our method on three benchmark datasets of MSR-VTT, MSVD and LSMDC, achieving new state-of-the-art results by up to 12% in relative improvement in Recall@1. Our findings thereby highlight the importance of joint text-video reasoning to extract important visual cues according to text. Full code and demo can be found at: https://layer6ai-labs.github.io/xpool/

Ziyi Wu, Nikita Dvornik, Klaus Greff et al. · gatech, nvidia

Understanding dynamics from visual observations is a challenging problem that requires disentangling individual objects from the scene and learning their interactions. While recent object-centric models can successfully decompose a scene into objects, modeling their dynamics effectively still remains a challenge. We address this problem by introducing SlotFormer -- a Transformer-based autoregressive model operating on learned object-centric representations. Given a video clip, our approach reasons over object features to model spatio-temporal relationships and predicts accurate future object states. In this paper, we successfully apply SlotFormer to perform video prediction on datasets with complex object interactions. Moreover, the unsupervised SlotFormer's dynamics model can be used to improve the performance on supervised downstream tasks, such as Visual Question Answering (VQA), and goal-conditioned planning. Compared to past works on dynamics modeling, our method achieves significantly better long-term synthesis of object dynamics, while retaining high quality visual generation. Besides, SlotFormer enables VQA models to reason about the future without object-level labels, even outperforming counterparts that use ground-truth annotations. Finally, we show its ability to serve as a world model for model-based planning, which is competitive with methods designed specifically for such tasks.

Yi Ru Wang, Yuchi Zhao, Haoping Xu et al. · gatech, nvidia

Transparent object perception is a crucial skill for applications such as robot manipulation in household and laboratory settings. Existing methods utilize RGB-D or stereo inputs to handle a subset of perception tasks including depth and pose estimation. However, transparent object perception remains to be an open problem. In this paper, we forgo the unreliable depth map from RGB-D sensors and extend the stereo based method. Our proposed method, MVTrans, is an end-to-end multi-view architecture with multiple perception capabilities, including depth estimation, segmentation, and pose estimation. Additionally, we establish a novel procedural photo-realistic dataset generation pipeline and create a large-scale transparent object detection dataset, Syn-TODD, which is suitable for training networks with all three modalities, RGB-D, stereo and multi-view RGB. Project Site: https://ac-rad.github.io/MVTrans/

Eric Heiden, Miles Macklin, Yashraj Narang et al. · gatech, nvidia

Robotic cutting of soft materials is critical for applications such as food processing, household automation, and surgical manipulation. As in other areas of robotics, simulators can facilitate controller verification, policy learning, and dataset generation. Moreover, differentiable simulators can enable gradient-based optimization, which is invaluable for calibrating simulation parameters and optimizing controllers. In this work, we present DiSECt: the first differentiable simulator for cutting soft materials. The simulator augments the finite element method (FEM) with a continuous contact model based on signed distance fields (SDF), as well as a continuous damage model that inserts springs on opposite sides of the cutting plane and allows them to weaken until zero stiffness, enabling crack formation. Through various experiments, we evaluate the performance of the simulator. We first show that the simulator can be calibrated to match resultant forces and deformation fields from a state-of-the-art commercial solver and real-world cutting datasets, with generality across cutting velocities and object instances. We then show that Bayesian inference can be performed efficiently by leveraging the differentiability of the simulator, estimating posteriors over hundreds of parameters in a fraction of the time of derivative-free methods. Next, we illustrate that control parameters in the simulation can be optimized to minimize cutting forces via lateral slicing motions. Finally, we conduct experiments on a real robot arm equipped with a slicing knife to infer simulation parameters from force measurements. By optimizing the slicing motion of the knife, we show on fruit cutting scenarios that the average knife force can be reduced by more than 40% compared to a vertical cutting motion. We publish code and additional materials on our project website at https://diff-cutting-sim.github.io.

Yun-Chun Chen, Haoda Li, Dylan Turpin et al. · gatech, nvidia

Learning to autonomously assemble shapes is a crucial skill for many robotic applications. While the majority of existing part assembly methods focus on correctly posing semantic parts to recreate a whole object, we interpret assembly more literally: as mating geometric parts together to achieve a snug fit. By focusing on shape alignment rather than semantic cues, we can achieve across-category generalization. In this paper, we introduce a novel task, pairwise 3D geometric shape mating, and propose Neural Shape Mating (NSM) to tackle this problem. Given the point clouds of two object parts of an unknown category, NSM learns to reason about the fit of the two parts and predict a pair of 3D poses that tightly mate them together. We couple the training of NSM with an implicit shape reconstruction task to make NSM more robust to imperfect point cloud observations. To train NSM, we present a self-supervised data collection pipeline that generates pairwise shape mating data with ground truth by randomly cutting an object mesh into two parts, resulting in a dataset that consists of 200K shape mating pairs from numerous object meshes with diverse cut types. We train NSM on the collected dataset and compare it with several point cloud registration methods and one part assembly baseline. Extensive experimental results and ablation studies under various settings demonstrate the effectiveness of the proposed algorithm. Additional material is available at: https://neural-shape-mating.github.io/

Silviu Pitis, Elliot Creager, Ajay Mandlekar et al. · gatech, nvidia

The number of states in a dynamic process is exponential in the number of objects, making reinforcement learning (RL) difficult in complex, multi-object domains. For agents to scale to the real world, they will need to react to and reason about unseen combinations of objects. We argue that the ability to recognize and use local factorization in transition dynamics is a key element in unlocking the power of multi-object reasoning. To this end, we show that (1) known local structure in the environment transitions is sufficient for an exponential reduction in the sample complexity of training a dynamics model, and (2) a locally factored dynamics model provably generalizes out-of-distribution to unseen states and actions. Knowing the local structure also allows us to predict which unseen states and actions this dynamics model will generalize to. We propose to leverage these observations in a novel Model-based Counterfactual Data Augmentation (MoCoDA) framework. MoCoDA applies a learned locally factored dynamics model to an augmented distribution of states and actions to generate counterfactual transitions for RL. MoCoDA works with a broader set of local structures than prior work and allows for direct control over the augmented training distribution. We show that MoCoDA enables RL agents to learn policies that generalize to unseen states and actions. We use MoCoDA to train an offline RL agent to solve an out-of-distribution robotics manipulation task on which standard offline RL algorithms fail.

Silvia Sellán, Yun-Chun Chen, Ziyi Wu et al. · gatech, nvidia

We introduce Breaking Bad, a large-scale dataset of fractured objects. Our dataset consists of over one million fractured objects simulated from ten thousand base models. The fracture simulation is powered by a recent physically based algorithm that efficiently generates a variety of fracture modes of an object. Existing shape assembly datasets decompose objects according to semantically meaningful parts, effectively modeling the construction process. In contrast, Breaking Bad models the destruction process of how a geometric object naturally breaks into fragments. Our dataset serves as a benchmark that enables the study of fractured object reassembly and presents new challenges for geometric shape understanding. We analyze our dataset with several geometry measurements and benchmark three state-of-the-art shape assembly deep learning methods under various settings. Extensive experimental results demonstrate the difficulty of our dataset, calling on future research in model designs specifically for the geometric shape assembly task. We host our dataset at https://breaking-bad-dataset.github.io/.

Lily Goli, Daniel Rebain, Sara Sabour et al. · gatech, nvidia

We introduce a technique for pairwise registration of neural fields that extends classical optimization-based local registration (i.e. ICP) to operate on Neural Radiance Fields (NeRF) -- neural 3D scene representations trained from collections of calibrated images. NeRF does not decompose illumination and color, so to make registration invariant to illumination, we introduce the concept of a ''surface field'' -- a field distilled from a pre-trained NeRF model that measures the likelihood of a point being on the surface of an object. We then cast nerf2nerf registration as a robust optimization that iteratively seeks a rigid transformation that aligns the surface fields of the two scenes. We evaluate the effectiveness of our technique by introducing a dataset of pre-trained NeRF scenes -- our synthetic scenes enable quantitative evaluations and comparisons to classical registration techniques, while our real scenes demonstrate the validity of our technique in real-world scenarios. Additional results available at: https://nerf2nerf.github.io

Nikita Dvornik, Isma Hadji, Ran Zhang et al. · gatech, nvidia

Instructional videos are an important resource to learn procedural tasks from human demonstrations. However, the instruction steps in such videos are typically short and sparse, with most of the video being irrelevant to the procedure. This motivates the need to temporally localize the instruction steps in such videos, i.e. the task called key-step localization. Traditional methods for key-step localization require video-level human annotations and thus do not scale to large datasets. In this work, we tackle the problem with no human supervision and introduce StepFormer, a self-supervised model that discovers and localizes instruction steps in a video. StepFormer is a transformer decoder that attends to the video with learnable queries, and produces a sequence of slots capturing the key-steps in the video. We train our system on a large dataset of instructional videos, using their automatically-generated subtitles as the only source of supervision. In particular, we supervise our system with a sequence of text narrations using an order-aware loss function that filters out irrelevant phrases. We show that our model outperforms all previous unsupervised and weakly-supervised approaches on step detection and localization by a large margin on three challenging benchmarks. Moreover, our model demonstrates an emergent property to solve zero-shot multi-step localization and outperforms all relevant baselines at this task.

Claas Voelcker, Victor Liao, Animesh Garg et al. · gatech, nvidia

Model-based reinforcement learning (MBRL) is a sample efficient technique to obtain control policies, yet unavoidable modeling errors often lead performance deterioration. The model in MBRL is often solely fitted to reconstruct dynamics, state observations in particular, while the impact of model error on the policy is not captured by the training objective. This leads to a mismatch between the intended goal of MBRL, enabling good policy and value learning, and the target of the loss function employed in practice, future state prediction. Naive intuition would suggest that value-aware model learning would fix this problem and, indeed, several solutions to this objective mismatch problem have been proposed based on theoretical analysis. However, they tend to be inferior in practice to commonly used maximum likelihood (MLE) based approaches. In this paper we propose the Value-gradient weighted Model Learning (VaGraM), a novel method for value-aware model learning which improves the performance of MBRL in challenging settings, such as small model capacity and the presence of distracting state dimensions. We analyze both MLE and value-aware approaches and demonstrate how they fail to account for exploration and the behavior of function approximation when learning value-aware models and highlight the additional goals that must be met to stabilize optimization in the deep learning setting. We verify our analysis by showing that our loss function is able to achieve high returns on the Mujoco benchmark suite while being more robust than maximum likelihood based approaches.

Yun-Chun Chen, Adithyavairavan Murali, Balakumar Sundaralingam et al. · gatech, nvidia

The pipeline of current robotic pick-and-place methods typically consists of several stages: grasp pose detection, finding inverse kinematic solutions for the detected poses, planning a collision-free trajectory, and then executing the open-loop trajectory to the grasp pose with a low-level tracking controller. While these grasping methods have shown good performance on grasping static objects on a table-top, the problem of grasping dynamic objects in constrained environments remains an open problem. We present Neural Motion Fields, a novel object representation which encodes both object point clouds and the relative task trajectories as an implicit value function parameterized by a neural network. This object-centric representation models a continuous distribution over the SE(3) space and allows us to perform grasping reactively by leveraging sampling-based MPC to optimize this value function.

Maria Attarian, Advaya Gupta, Ziyi Zhou et al. · gatech, nvidia

Cognitive planning is the structural decomposition of complex tasks into a sequence of future behaviors. In the computational setting, performing cognitive planning entails grounding plans and concepts in one or more modalities in order to leverage them for low level control. Since real-world tasks are often described in natural language, we devise a cognitive planning algorithm via language-guided video prediction. Current video prediction models do not support conditioning on natural language instructions. Therefore, we propose a new video prediction architecture which leverages the power of pre-trained transformers.The network is endowed with the ability to ground concepts based on natural language input with generalization to unseen objects. We demonstrate the effectiveness of this approach on a new simulation dataset, where each task is defined by a high-level action described in natural language. Our experiments compare our method again stone video generation baseline without planning or action grounding and showcase significant improvements. Our ablation studies highlight an improved generalization to unseen objects that natural language embeddings offer to concept grounding ability, as well as the importance of planning towards visual "imagination" of a task.

Mohan Zhang, Xiaozhou Wang, Benjamin Decardi-Nelson et al. · gatech, nvidia

Traditional biological and pharmaceutical manufacturing plants are controlled by human workers or pre-defined thresholds. Modernized factories have advanced process control algorithms such as model predictive control (MPC). However, there is little exploration of applying deep reinforcement learning to control manufacturing plants. One of the reasons is the lack of high fidelity simulations and standard APIs for benchmarking. To bridge this gap, we develop an easy-to-use library that includes five high-fidelity simulation environments: BeerFMTEnv, ReactorEnv, AtropineEnv, PenSimEnv and mAbEnv, which cover a wide range of manufacturing processes. We build these environments on published dynamics models. Furthermore, we benchmark online and offline, model-based and model-free reinforcement learning algorithms for comparisons of follow-up research.

Zihan Zhou, Animesh Garg · gatech, nvidia

We propose Structured Exploration with Achievements (SEA), a multi-stage reinforcement learning algorithm designed for achievement-based environments, a particular type of environment with an internal achievement set. SEA first uses offline data to learn a representation of the known achievements with a determinant loss function, then recovers the dependency graph of the learned achievements with a heuristic algorithm, and finally interacts with the environment online to learn policies that master known achievements and explore new ones with a controller built with the recovered dependency graph. We empirically demonstrate that SEA can recover the achievement structure accurately and improve exploration in hard domains such as Crafter that are procedurally generated with high-dimensional observations like images.

Chaitanya Devaguptapu, Samarth Sinha, K J Joseph et al. · gatech, nvidia

Models pre-trained on large-scale datasets are often fine-tuned to support newer tasks and datasets that arrive over time. This process necessitates storing copies of the model over time for each task that the pre-trained model is fine-tuned to. Building on top of recent model patching work, we propose $Δ$-Patching for fine-tuning neural network models in an efficient manner, without the need to store model copies. We propose a simple and lightweight method called $Δ$-Networks to achieve this objective. Our comprehensive experiments across setting and architecture variants show that $Δ$-Networks outperform earlier model patching work while only requiring a fraction of parameters to be trained. We also show that this approach can be used for other problem settings such as transfer learning and zero-shot domain adaptation, as well as other tasks such as detection and segmentation.

Riashat Islam, Samarth Sinha, Homanga Bharadhwaj et al. · gatech, mila

Learning policies from fixed offline datasets is a key challenge to scale up reinforcement learning (RL) algorithms towards practical applications. This is often because off-policy RL algorithms suffer from distributional shift, due to mismatch between dataset and the target policy, leading to high variance and over-estimation of value functions. In this work, we propose variance regularization for offline RL algorithms, using stationary distribution corrections. We show that by using Fenchel duality, we can avoid double sampling issues for computing the gradient of the variance regularizer. The proposed algorithm for offline variance regularization (OVAR) can be used to augment any existing offline policy optimization algorithms. We show that the regularizer leads to a lower bound to the offline policy optimization objective, which can help avoid over-estimation errors, and explains the benefits of our approach across a range of continuous control domains when compared to existing state-of-the-art algorithms.

Shutong Zhang, Yi-Ling Qiao, Guanglei Zhu et al. · cmu

Various heuristic objectives for modeling hand-object interaction have been proposed in past work. However, due to the lack of a cohesive framework, these objectives often possess a narrow scope of applicability and are limited by their efficiency or accuracy. In this paper, we propose HandyPriors, a unified and general pipeline for pose estimation in human-object interaction scenes by leveraging recent advances in differentiable physics and rendering. Our approach employs rendering priors to align with input images and segmentation masks along with physics priors to mitigate penetration and relative-sliding across frames. Furthermore, we present two alternatives for hand and object pose estimation. The optimization-based pose estimation achieves higher accuracy, while the filtering-based tracking, which utilizes the differentiable priors as dynamics and observation models, executes faster. We demonstrate that HandyPriors attains comparable or superior results in the pose estimation task, and that the differentiable physics module can predict contact information for pose refinement. We also show that our approach generalizes to perception tasks, including robotic hand manipulation and human-object pose estimation in the wild.

Wei Yu, Runjia Qian, Yumeng Li et al. · amazon-science

Video diffusion models are moving beyond short, plausible clips toward world simulators that must remain consistent under camera motion, revisits, and intervention. Yet spatial memory remains a key bottleneck: explicit 3D structures can improve reprojection-based consistency but struggle to depict moving objects, while implicit memory often produces inaccurate camera motion even with correct poses. We propose Mosaic Memory (MosaicMem), a hybrid spatial memory that lifts patches into 3D for reliable localization and targeted retrieval, while exploiting the model's native conditioning to preserve prompt-following generation. MosaicMem composes spatially aligned patches in the queried view via a patch-and-compose interface, preserving what should persist while allowing the model to inpaint what should evolve. With PRoPE camera conditioning and two new memory alignment methods, experiments show improved pose adherence compared to implicit memory and stronger dynamic modeling than explicit baselines. MosaicMem further enables minute-level navigation, memory-based scene editing, and autoregressive rollout.

Masoud Moghani, Mahdi Azizian, Animesh Garg et al.

Large-scale robot datasets have facilitated the learning of a wide range of robot manipulation skills, but these datasets remain difficult to collect and scale further, owing to the intractable amount of human time, effort, and cost required. Simulation and synthetic data generation have proven to be an effective alternative to fuel this need for data, especially with the advent of recent work showing that such synthetic datasets can dramatically reduce real-world data requirements and facilitate generalization to novel scenarios unseen in real-world demonstrations. However, this paradigm has been limited to rigid-body tasks, which are easy to simulate. Deformable object manipulation encompasses a large portion of real-world manipulation and remains a crucial gap to address towards increasing adoption of the synthetic simulation data paradigm. In this paper, we introduce SoftMimicGen, an automated data generation pipeline for deformable object manipulation tasks. We introduce a suite of high-fidelity simulation environments that encompasses a wide range of deformable objects (stuffed animal, rope, tissue, towel) and manipulation behaviors (high-precision threading, dynamic whipping, folding, pick-and-place), across four robot embodiments: a single-arm manipulator, bimanual arms, a humanoid, and a surgical robot. We apply SoftMimicGen to generate datasets across the task suite, train high-performing policies from the data, and systematically analyze the data generation system. Project website: \href{https://softmimicgen.github.io}{softmimicgen.github.io}.

Ignat Georgiev, Varun Giridhar, Nicklas Hansen et al.

Reinforcement Learning (RL) has made significant strides in complex tasks but struggles in multi-task settings with different embodiments. World model methods offer scalability by learning a simulation of the environment but often rely on inefficient gradient-free optimization methods for policy extraction. In contrast, gradient-based methods exhibit lower variance but fail to handle discontinuities. Our work reveals that well-regularized world models can generate smoother optimization landscapes than the actual dynamics, facilitating more effective first-order optimization. We introduce Policy learning with multi-task World Models (PWM), a novel model-based RL algorithm for continuous control. Initially, the world model is pre-trained on offline data, and then policies are extracted from it using first-order optimization in less than 10 minutes per task. PWM effectively solves tasks with up to 152 action dimensions and outperforms methods that use ground-truth dynamics. Additionally, PWM scales to an 80-task setting, achieving up to 27% higher rewards than existing baselines without relying on costly online planning. Visualizations and code are available at https://www.imgeorgiev.com/pwm/.

Zihan Zhou, Animesh Garg, Ajay Mandlekar et al.

Long-horizon manipulation has been a long-standing challenge in the robotics community. We propose ReinforceGen, a system that combines task decomposition, data generation, imitation learning, and motion planning to form an initial solution, and improves each component through reinforcement-learning-based fine-tuning. ReinforceGen first segments the task into multiple localized skills, which are connected through motion planning. The skills and motion planning targets are trained with imitation learning on a dataset generated from 10 human demonstrations, and then fine-tuned through online adaptation and reinforcement learning. When benchmarked on the Robosuite dataset, ReinforceGen reaches 80% success rate on all tasks with visuomotor controls in the highest reset range setting. Additional ablation studies show that our fine-tuning approaches contributes to an 89% average performance increase. More results and videos available in https://reinforcegen.github.io/

Nitya Thakkar, Mert Yuksekgonul, Jake Silberg et al. · stanford

Peer review at AI conferences is stressed by rapidly rising submission volumes, leading to deteriorating review quality and increased author dissatisfaction. To address these issues, we developed Review Feedback Agent, a system leveraging multiple large language models (LLMs) to improve review clarity and actionability by providing automated feedback on vague comments, content misunderstandings, and unprofessional remarks to reviewers. Implemented at ICLR 2025 as a large randomized control study, our system provided optional feedback to more than 20,000 randomly selected reviews. To ensure high-quality feedback for reviewers at this scale, we also developed a suite of automated reliability tests powered by LLMs that acted as guardrails to ensure feedback quality, with feedback only being sent to reviewers if it passed all the tests. The results show that 27% of reviewers who received feedback updated their reviews, and over 12,000 feedback suggestions from the agent were incorporated by those reviewers. This suggests that many reviewers found the AI-generated feedback sufficiently helpful to merit updating their reviews. Incorporating AI feedback led to significantly longer reviews (an average increase of 80 words among those who updated after receiving feedback) and more informative reviews, as evaluated by blinded researchers. Moreover, reviewers who were selected to receive AI feedback were also more engaged during paper rebuttals, as seen in longer author-reviewer discussions. This work demonstrates that carefully designed LLM-generated review feedback can enhance peer review quality by making reviews more specific and actionable while increasing engagement between reviewers and authors. The Review Feedback Agent is publicly available at https://github.com/zou-group/review_feedback_agent.

Ayush Agarwal, Ansh Gandhi, Jeremy A. Collins et al.

The scarcity of large-scale, high-quality demonstration data remains a bottleneck in scaling imitation learning for robotic manipulation. We present COBALT, a teleoperation platform designed to democratize robot learning at scale both in simulation and in the real world. By leveraging vectorized environments, our scalable, load-balanced infrastructure supports concurrent teleoperation by multiple users on a single GPU, yielding a significant reduction in teleoperation cost. Operators can connect from nearly anywhere on Earth using commonly available devices, including single or dual smartphones, VR headsets, 3D mice, and keyboards. An inmemory data cache and efficient video streaming keep control and rendering synchronous, sustaining dozens of concurrent users at 20 Hz with sub-100 ms end-to-end latency for up to 8 concurrent users per GPU. We also demonstrate stable operation supporting 256 simulated clients across 8 GPUs, underscoring the system's ability to scale across hardware and within individual servers. We perform a comprehensive user study showing that phone-based teleoperation performs comparably to or better than specialized hardware, enabling faster, more ergonomic data collection. To ensure data quality, COBALT logs a suite of real-time metrics to automatically filter suboptimal demonstrations. We further demonstrate that a structured user training curriculum significantly improves data collection quality. Guided by insights from our user study, we crowdsource the collection of a large-scale, high-quality pilot dataset with 7500+ demonstrations (50+ hours) collected with smartphones across nine countries over five days. We validate the dataset's quality by training state-of-the-art imitation learning algorithms. Please visit \href{https://cobalt-teleop.github.io/}{cobalt-teleop.github.io} for more details.

Zihan Zhou, Jonathan Booher, Khashayar Rohanimanesh et al.

Safe reinforcement learning tasks are a challenging domain despite being very common in the real world. The widely adopted CMDP model constrains the risks in expectation, which makes room for dangerous behaviors in long-tail states. In safety-critical domains, such behaviors could lead to disastrous outcomes. To address this issue, we first describe the problem with a stronger Uniformly Constrained MDP (UCMDP) model where we impose constraints on all reachable states; we then propose Objective Suppression, a novel method that adaptively suppresses the task reward maximizing objectives according to a safety critic, as a solution to the Lagrangian dual of a UCMDP. We benchmark Objective Suppression in two multi-constraint safety domains, including an autonomous driving domain where any incorrect behavior can lead to disastrous consequences. On the driving domain, we evaluate on open source and proprietary data and evaluate transfer to a real autonomous fleet. Empirically, we demonstrate that our proposed method, when combined with existing safe RL algorithms, can match the task reward achieved by baselines with significantly fewer constraint violations.

Panteha Naderian, Gabriel Loaiza-Ganem, Harry J. Braviner et al.

Multi-goal reaching is an important problem in reinforcement learning needed to achieve algorithmic generalization. Despite recent advances in this field, current algorithms suffer from three major challenges: high sample complexity, learning only a single way of reaching the goals, and difficulties in solving complex motion planning tasks. In order to address these limitations, we introduce the concept of cumulative accessibility functions, which measure the reachability of a goal from a given state within a specified horizon. We show that these functions obey a recurrence relation, which enables learning from offline interactions. We also prove that optimal cumulative accessibility functions are monotonic in the planning horizon. Additionally, our method can trade off speed and reliability in goal-reaching by suggesting multiple paths to a single goal depending on the provided horizon. We evaluate our approach on a set of multi-goal discrete and continuous control tasks. We show that our method outperforms state-of-the-art goal-reaching algorithms in success rate, sample complexity, and path optimality. Our code is available at https://github.com/layer6ai-labs/CAE, and additional visualizations can be found at https://sites.google.com/view/learning-cae/.

Samarth Sinha, Homanga Bharadhwaj, Anirudh Goyal et al.

Although deep learning models have achieved state-of-the-art performance on a number of vision tasks, generalization over high dimensional multi-modal data, and reliable predictive uncertainty estimation are still active areas of research. Bayesian approaches including Bayesian Neural Nets (BNNs) do not scale well to modern computer vision tasks, as they are difficult to train, and have poor generalization under dataset-shift. This motivates the need for effective ensembles which can generalize and give reliable uncertainty estimates. In this paper, we target the problem of generating effective ensembles of neural networks by encouraging diversity in prediction. We explicitly optimize a diversity inducing adversarial loss for learning the stochastic latent variables and thereby obtain diversity in the output predictions necessary for modeling multi-modal data. We evaluate our method on benchmark datasets: MNIST, CIFAR100, TinyImageNet and MIT Places 2, and compared to the most competitive baselines show significant improvements in classification accuracy, under a shift in the data distribution and in out-of-distribution detection. Code will be released in this url https://github.com/rvl-lab-utoronto/dibs

Vinu Joseph, Saurav Muralidharan, Animesh Garg et al.

Deep neural networks (DNNs) frequently contain far more weights, represented at a higher precision, than are required for the specific task which they are trained to perform. Consequently, they can often be compressed using techniques such as weight pruning and quantization that reduce both the model size and inference time without appreciable loss in accuracy. However, finding the best compression strategy and corresponding target sparsity for a given DNN, hardware platform, and optimization objective currently requires expensive, frequently manual, trial-and-error experimentation. In this paper, we introduce a programmable system for model compression called Condensa. Users programmatically compose simple operators, in Python, to build more complex and practically interesting compression strategies. Given a strategy and user-provided objective (such as minimization of running time), Condensa uses a novel Bayesian optimization-based algorithm to automatically infer desirable sparsities. Our experiments on four real-world DNNs demonstrate memory footprint and hardware runtime throughput improvements of 188x and 2.59x, respectively, using at most ten samples per search. We have released a reference implementation of Condensa at https://github.com/NVlabs/condensa.

Kourosh Darvish, Marta Skreta, Yuchi Zhao et al. · gatech, nvidia

Chemistry experiments can be resource- and labor-intensive, often requiring manual tasks like polishing electrodes in electrochemistry. Traditional lab automation infrastructure faces challenges adapting to new experiments. To address this, we introduce ORGANA, an assistive robotic system that automates diverse chemistry experiments using decision-making and perception tools. It makes decisions with chemists in the loop to control robots and lab devices. ORGANA interacts with chemists using Large Language Models (LLMs) to derive experiment goals, handle disambiguation, and provide experiment logs. ORGANA plans and executes complex tasks with visual feedback, while supporting scheduling and parallel task execution. We demonstrate ORGANA's capabilities in solubility, pH measurement, recrystallization, and electrochemistry experiments. In electrochemistry, it executes a 19-step plan in parallel to characterize quinone derivatives for flow batteries. Our user study shows ORGANA reduces frustration and physical demand by over 50%, with users saving an average of 80.3% of their time when using it.

Zihan Zhou, Animesh Garg, Dieter Fox et al. · mit

Robot learning has proven to be a general and effective technique for programming manipulators. Imitation learning is able to teach robots solely from human demonstrations but is bottlenecked by the capabilities of the demonstrations. Reinforcement learning uses exploration to discover better behaviors; however, the space of possible improvements can be too large to start from scratch. And for both techniques, the learning difficulty increases proportional to the length of the manipulation task. Accounting for this, we propose SPIRE, a system that first uses Task and Motion Planning (TAMP) to decompose tasks into smaller learning subproblems and second combines imitation and reinforcement learning to maximize their strengths. We develop novel strategies to train learning agents when deployed in the context of a planning system. We evaluate SPIRE on a suite of long-horizon and contact-rich robot manipulation problems. We find that SPIRE outperforms prior approaches that integrate imitation learning, reinforcement learning, and planning by 35% to 50% in average task performance, is 6 times more data efficient in the number of human demonstrations needed to train proficient agents, and learns to complete tasks nearly twice as efficiently. View https://sites.google.com/view/spire-corl-2024 for more details.

Andrew Melnik, Robin Schiewer, Moritz Lange et al.

Physical reasoning is a crucial aspect in the development of general AI systems, given that human learning starts with interacting with the physical world before progressing to more complex concepts. Although researchers have studied and assessed the physical reasoning of AI approaches through various specific benchmarks, there is no comprehensive approach to evaluating and measuring progress. Therefore, we aim to offer an overview of existing benchmarks and their solution approaches and propose a unified perspective for measuring the physical reasoning capacity of AI systems. We select benchmarks that are designed to test algorithmic performance in physical reasoning tasks. While each of the selected benchmarks poses a unique challenge, their ensemble provides a comprehensive proving ground for an AI generalist agent with a measurable skill level for various physical reasoning concepts. This gives an advantage to such an ensemble of benchmarks over other holistic benchmarks that aim to simulate the real world by intertwining its complexity and many concepts. We group the presented set of physical reasoning benchmarks into subcategories so that more narrow generalist AI agents can be tested first on these groups.

Ishika Singh, Ankit Goyal, Stan Birchfield et al. · nvidia

We introduce OG-VLA, a novel architecture and learning framework that combines the generalization strengths of Vision Language Action models (VLAs) with the robustness of 3D-aware policies. We address the challenge of mapping natural language instructions and one or more RGBD observations to quasi-static robot actions. 3D-aware robot policies achieve state-of-the-art performance on precise robot manipulation tasks, but struggle with generalization to unseen instructions, scenes, and objects. On the other hand, VLAs excel at generalizing across instructions and scenes, but can be sensitive to camera and robot pose variations. We leverage prior knowledge embedded in language and vision foundation models to improve generalization of 3D-aware keyframe policies. OG-VLA unprojects input observations from diverse views into a point cloud which is then rendered from canonical orthographic views, ensuring input view invariance and consistency between input and output spaces. These canonical views are processed with a vision backbone, a Large Language Model (LLM), and an image diffusion model to generate images that encode the next position and orientation of the end-effector on the input scene. Evaluations on the Arnold and Colosseum benchmarks demonstrate state-of-the-art generalization to unseen environments, with over 40% relative improvements while maintaining robust performance in seen settings. We also show real-world adaption in 3 to 5 demonstrations along with strong generalization. Videos and resources at https://og-vla.github.io/

Yuchi Zhao, Miroslav Bogdanovic, Chengyuan Luo et al.

Object placement in robotic tasks is inherently challenging due to the diversity of object geometries and placement configurations. To address this, we propose AnyPlace, a two-stage method trained entirely on synthetic data, capable of predicting a wide range of feasible placement poses for real-world tasks. Our key insight is that by leveraging a Vision-Language Model (VLM) to identify rough placement locations, we focus only on the relevant regions for local placement, which enables us to train the low-level placement-pose-prediction model to capture diverse placements efficiently. For training, we generate a fully synthetic dataset of randomly generated objects in different placement configurations (insertion, stacking, hanging) and train local placement-prediction models. We conduct extensive evaluations in simulation, demonstrating that our method outperforms baselines in terms of success rate, coverage of possible placement modes, and precision. In real-world experiments, we show how our approach directly transfers models trained purely on synthetic data to the real world, where it successfully performs placements in scenarios where other models struggle -- such as with varying object geometries, diverse placement modes, and achieving high precision for fine placement. More at: https://any-place.github.io.

Ezra Ameperosa, Jeremy A. Collins, Mrinal Jain et al.

Imitation learning in robotics faces significant challenges in generalization due to the complexity of robotic environments and the high cost of data collection. We introduce RoCoDA, a novel method that unifies the concepts of invariance, equivariance, and causality within a single framework to enhance data augmentation for imitation learning. RoCoDA leverages causal invariance by modifying task-irrelevant subsets of the environment state without affecting the policy's output. Simultaneously, we exploit SE(3) equivariance by applying rigid body transformations to object poses and adjusting corresponding actions to generate synthetic demonstrations. We validate RoCoDA through extensive experiments on five robotic manipulation tasks, demonstrating improvements in policy performance, generalization, and sample efficiency compared to state-of-the-art data augmentation methods. Our policies exhibit robust generalization to unseen object poses, textures, and the presence of distractors. Furthermore, we observe emergent behavior such as re-grasping, indicating policies trained with RoCoDA possess a deeper understanding of task dynamics. By leveraging invariance, equivariance, and causality, RoCoDA provides a principled approach to data augmentation in imitation learning, bridging the gap between geometric symmetries and causal reasoning. Project Page: https://rocoda.github.io

Pengsong Zhang, Heng Zhang, Huazhe Xu et al.

Scientific discovery is poised for rapid advancement through advanced robotics and artificial intelligence. Current scientific practices face substantial limitations as manual experimentation remains time-consuming and resource-intensive, while multidisciplinary research demands knowledge integration beyond individual researchers' expertise boundaries. Here, we envision an autonomous generalist scientist (AGS) concept combines agentic AI and embodied robotics to automate the entire research lifecycle. This system could dynamically interact with both physical and virtual environments while facilitating the integration of knowledge across diverse scientific disciplines. By deploying these technologies throughout every research stage -- spanning literature review, hypothesis generation, experimentation, and manuscript writing -- and incorporating internal reflection alongside external feedback, this system aims to significantly reduce the time and resources needed for scientific discovery. Building on the evolution from virtual AI scientists to versatile generalist AI-based robot scientists, AGS promises groundbreaking potential. As these autonomous systems become increasingly integrated into the research process, we hypothesize that scientific discovery might adhere to new scaling laws, potentially shaped by the number and capabilities of these autonomous systems, offering novel perspectives on how knowledge is generated and evolves. The adaptability of embodied robots to extreme environments, paired with the flywheel effect of accumulating scientific knowledge, holds the promise of continually pushing beyond both physical and intellectual frontiers.

Albert Wilcox, Mohamed Ghanem, Masoud Moghani et al.

Imitation Learning can train robots to perform complex and diverse manipulation tasks, but learned policies are brittle with observations outside of the training distribution. 3D scene representations that incorporate observations from calibrated RGBD cameras have been proposed as a way to mitigate this, but in our evaluations with unseen embodiments and camera viewpoints they show only modest improvement. To address those challenges, we propose Adapt3R, a general-purpose 3D observation encoder which synthesizes data from calibrated RGBD cameras into a vector that can be used as conditioning for arbitrary IL algorithms. The key idea is to use a pretrained 2D backbone to extract semantic information, using 3D only as a medium to localize this information with respect to the end-effector. We show across 93 simulated and 6 real tasks that when trained end-to-end with a variety of IL algorithms, Adapt3R maintains these algorithms' learning capacity while enabling zero-shot transfer to novel embodiments and camera poses.

Walker Byrnes, Miroslav Bogdanovic, Avi Balakirsky et al.

Intelligent and reliable task planning is a core capability for generalized robotics, requiring a descriptive domain representation that sufficiently models all object and state information for the scene. We present CLIMB, a continual learning framework for robot task planning that leverages foundation models and execution feedback to guide domain model construction. CLIMB can build a model from a natural language description, learn non-obvious predicates while solving tasks, and store that information for future problems. We demonstrate the ability of CLIMB to improve performance in common planning environments compared to baseline methods. We also develop the BlocksWorld++ domain, a simulated environment with an easily usable real counterpart, together with a curriculum of tasks with progressing difficulty for evaluating continual learning. Additional details and demonstrations for this system can be found at https://plan-with-climb.github.io/ .

Joshua Bagajo, Clemens Schwarke, Victor Klemm et al.

Differentiable simulators provide analytic gradients, enabling more sample-efficient learning algorithms and paving the way for data intensive learning tasks such as learning from images. In this work, we demonstrate that locomotion policies trained with analytic gradients from a differentiable simulator can be successfully transferred to the real world. Typically, simulators that offer informative gradients lack the physical accuracy needed for sim-to-real transfer, and vice-versa. A key factor in our success is a smooth contact model that combines informative gradients with physical accuracy, ensuring effective transfer of learned behaviors. To the best of our knowledge, this is the first time a real quadrupedal robot is able to locomote after training exclusively in a differentiable simulation.

Jeremy A. Collins, Loránd Cheng, Kunal Aneja et al.

Action-labeled data for robotics is scarce and expensive, limiting the generalization of learned policies. In contrast, vast amounts of action-free video data are readily available, but translating these observations into effective policies remains a challenge. We introduce AMPLIFY, a novel framework that leverages large-scale video data by encoding visual dynamics into compact, discrete motion tokens derived from keypoint trajectories. Our modular approach separates visual motion prediction from action inference, decoupling the challenges of learning what motion defines a task from how robots can perform it. We train a forward dynamics model on abundant action-free videos and an inverse dynamics model on a limited set of action-labeled examples, allowing for independent scaling. Extensive evaluations demonstrate that the learned dynamics are both accurate, achieving up to 3.7x better MSE and over 2.5x better pixel prediction accuracy compared to prior approaches, and broadly useful. In downstream policy learning, our dynamics predictions enable a 1.2-2.2x improvement in low-data regimes, a 1.4x average improvement by learning from action-free human videos, and the first generalization to LIBERO tasks from zero in-distribution action data. Beyond robotic control, we find the dynamics learned by AMPLIFY to be a versatile latent world model, enhancing video prediction quality. Our results present a novel paradigm leveraging heterogeneous data sources to build efficient, generalizable world models. More information can be found at https://amplify-robotics.github.io/.

Weiyu Liu, Jiayuan Mao, Joy Hsu et al.

In this paper, we propose composable part-based manipulation (CPM), a novel approach that leverages object-part decomposition and part-part correspondences to improve learning and generalization of robotic manipulation skills. By considering the functional correspondences between object parts, we conceptualize functional actions, such as pouring and constrained placing, as combinations of different correspondence constraints. CPM comprises a collection of composable diffusion models, where each model captures a different inter-object correspondence. These diffusion models can generate parameters for manipulation skills based on the specific object parts. Leveraging part-based correspondences coupled with the task decomposition into distinct constraints enables strong generalization to novel objects and object categories. We validate our approach in both simulated and real-world scenarios, demonstrating its effectiveness in achieving robust and generalized manipulation capabilities.

Tongzhou Mu, Yijie Guo, Jie Xu et al. · nvidia

Encouraged by the remarkable achievements of language and vision foundation models, developing generalist robotic agents through imitation learning, using large demonstration datasets, has become a prominent area of interest in robot learning. The efficacy of imitation learning is heavily reliant on the quantity and quality of the demonstration datasets. In this study, we aim to scale up demonstrations in a data-efficient way to facilitate the learning of generalist robotic agents. We introduce AdaDemo (Adaptive Online Demonstration Expansion), a general framework designed to improve multi-task policy learning by actively and continually expanding the demonstration dataset. AdaDemo strategically collects new demonstrations to address the identified weakness in the existing policy, ensuring data efficiency is maximized. Through a comprehensive evaluation on a total of 22 tasks across two robotic manipulation benchmarks (RLBench and Adroit), we demonstrate AdaDemo's capability to progressively improve policy performance by guiding the generation of high-quality demonstration datasets in a data-efficient manner.

Liquan Wang, Nikita Dvornik, Rafael Dubeau et al.

When humans perform a task with an articulated object, they interact with the object only in a handful of ways, while the space of all possible interactions is nearly endless. This is because humans have prior knowledge about what interactions are likely to be successful, i.e., to open a new door we first try the handle. While learning such priors without supervision is easy for humans, it is notoriously hard for machines. In this work, we tackle unsupervised learning of priors of useful interactions with articulated objects, which we call interaction modes. In contrast to the prior art, we use no supervision or privileged information; we only assume access to the depth sensor in the simulator to learn the interaction modes. More precisely, we define a successful interaction as the one changing the visual environment substantially and learn a generative model of such interactions, that can be conditioned on the desired goal state of the object. In our experiments, we show that our model covers most of the human interaction modes, outperforms existing state-of-the-art methods for affordance learning, and can generalize to objects never seen during training. Additionally, we show promising results in the goal-conditional setup, where our model can be quickly fine-tuned to perform a given task. We show in the experiments that such affordance learning predicts interaction which covers most modes of interaction for the querying articulated object and can be fine-tuned to a goal-conditional model. For supplementary: https://actaim.github.io.

Ziyi Wu, Jingyu Hu, Wuyue Lu et al.

Object-centric learning aims to represent visual data with a set of object entities (a.k.a. slots), providing structured representations that enable systematic generalization. Leveraging advanced architectures like Transformers, recent approaches have made significant progress in unsupervised object discovery. In addition, slot-based representations hold great potential for generative modeling, such as controllable image generation and object manipulation in image editing. However, current slot-based methods often produce blurry images and distorted objects, exhibiting poor generative modeling capabilities. In this paper, we focus on improving slot-to-image decoding, a crucial aspect for high-quality visual generation. We introduce SlotDiffusion -- an object-centric Latent Diffusion Model (LDM) designed for both image and video data. Thanks to the powerful modeling capacity of LDMs, SlotDiffusion surpasses previous slot models in unsupervised object segmentation and visual generation across six datasets. Furthermore, our learned object features can be utilized by existing object-centric dynamics models, improving video prediction quality and downstream temporal reasoning tasks. Finally, we demonstrate the scalability of SlotDiffusion to unconstrained real-world datasets such as PASCAL VOC and COCO, when integrated with self-supervised pre-trained image encoders.

Chenjia Bai, Lingxiao Wang, Zhuoran Yang et al.

Offline Reinforcement Learning (RL) aims to learn policies from previously collected datasets without exploring the environment. Directly applying off-policy algorithms to offline RL usually fails due to the extrapolation error caused by the out-of-distribution (OOD) actions. Previous methods tackle such problem by penalizing the Q-values of OOD actions or constraining the trained policy to be close to the behavior policy. Nevertheless, such methods typically prevent the generalization of value functions beyond the offline data and also lack precise characterization of OOD data. In this paper, we propose Pessimistic Bootstrapping for offline RL (PBRL), a purely uncertainty-driven offline algorithm without explicit policy constraints. Specifically, PBRL conducts uncertainty quantification via the disagreement of bootstrapped Q-functions, and performs pessimistic updates by penalizing the value function based on the estimated uncertainty. To tackle the extrapolating error, we further propose a novel OOD sampling method. We show that such OOD sampling and pessimistic bootstrapping yields provable uncertainty quantifier in linear MDPs, thus providing the theoretical underpinning for PBRL. Extensive experiments on D4RL benchmark show that PBRL has better performance compared to the state-of-the-art algorithms.

Matthias Weissenbacher, Samarth Sinha, Animesh Garg et al.

Offline reinforcement learning leverages large datasets to train policies without interactions with the environment. The learned policies may then be deployed in real-world settings where interactions are costly or dangerous. Current algorithms over-fit to the training dataset and as a consequence perform poorly when deployed to out-of-distribution generalizations of the environment. We aim to address these limitations by learning a Koopman latent representation which allows us to infer symmetries of the system's underlying dynamic. The latter is then utilized to extend the otherwise static offline dataset during training; this constitutes a novel data augmentation framework which reflects the system's dynamic and is thus to be interpreted as an exploration of the environments phase space. To obtain the symmetries we employ Koopman theory in which nonlinear dynamics are represented in terms of a linear operator acting on the space of measurement functions of the system and thus symmetries of the dynamics may be inferred directly. We provide novel theoretical results on the existence and nature of symmetries relevant for control systems such as reinforcement learning settings. Moreover, we empirically evaluate our method on several benchmark offline reinforcement learning tasks and datasets including D4RL, Metaworld and Robosuite and find that by using our framework we consistently improve the state-of-the-art of model-free Q-learning methods.

Matthew S. Zhang, Murat A. Erdogdu, Animesh Garg

Policy gradient methods have been frequently applied to problems in control and reinforcement learning with great success, yet existing convergence analysis still relies on non-intuitive, impractical and often opaque conditions. In particular, existing rates are achieved in limited settings, under strict regularity conditions. In this work, we establish explicit convergence rates of policy gradient methods, extending the convergence regime to weakly smooth policy classes with $L_2$ integrable gradient. We provide intuitive examples to illustrate the insight behind these new conditions. Notably, our analysis also shows that convergence rates are achievable for both the standard policy gradient and the natural policy gradient algorithms under these assumptions. Lastly we provide performance guarantees for the converged policies.