Creating 3D graphics content for immersive and interactive worlds remains labor-intensive, limiting our ability to create large-scale synthetic data for training foundation models. Recent methods aim to alleviate this, but they often focus on a single aspect (e.g., layout) and do not improve generation quality by simply scaling computational resources. We recast 3D environment generation as a sequential decision-making problem, using Vision-Language Models (VLMs) as policies that output actions to jointly craft a 3D environment's layout, materials, lighting, and assets. Our framework, 3D-Generalist, trains VLMs to generate more prompt-aligned 3D environments via self-improvement fine-tuning. We demonstrate the effectiveness of 3D-Generalist and our training strategy in generating simulation-ready 3D environments. We also demonstrate its quality and scalability for synthetic data generation by pretraining a vision foundation model on the generated data. After fine-tuning on downstream tasks, we show that it surpasses models pre-trained on meticulously human-crafted synthetic data and approaches results achieved when training with orders of magnitude larger real data.

Generative models such as diffusion models, excel at capturing high-dimensional distributions with diverse input modalities, e.g. robot trajectories, but are less effective at multistep constraint reasoning. Task and Motion Planning (TAMP) approaches are suited for planning multi-step autonomous robot manipulation. However, it can be difficult to apply them to domains where the environment and its dynamics are not fully known. We propose to overcome these limitations by composing diffusion models using a TAMP system. We use the learned components for constraints and samplers that are difficult to engineer in the planning model, and use a TAMP solver to search for the task plan with constraint-satisfying action parameter values. To tractably make predictions for unseen objects in the environment, we define the learned samplers and TAMP operators on learned latent embedding of changing object states. We evaluate our approach in a simulated articulated object manipulation domain and show how the combination of classical TAMP, generative modeling, and latent embedding enables multi-step constraint-based reasoning. We also apply the learned sampler in the real world. Website: https://sites.google.com/view/dimsam-tamp.

Extrinsic manipulation, the use of environment contacts to achieve manipulation objectives, enables strategies that are otherwise impossible with a parallel jaw gripper. However, orchestrating a long-horizon sequence of contact interactions between the robot, object, and environment is notoriously challenging due to the scene diversity, large action space, and difficult contact dynamics. We observe that most extrinsic manipulation are combinations of short-horizon primitives, each of which depend strongly on initializing from a desirable contact configuration to succeed. Therefore, we propose to generalize one extrinsic manipulation trajectory to diverse objects and environments by retargeting contact requirements. We prepare a single library of robust short-horizon, goal-conditioned primitive policies, and design a framework to compose state constraints stemming from contacts specifications of each primitive. Given a test scene and a single demo prescribing the primitive sequence, our method enforces the state constraints on the test scene and find intermediate goal states using inverse kinematics. The goals are then tracked by the primitive policies. Using a 7+1 DoF robotic arm-gripper system, we achieved an overall success rate of 80.5% on hardware over 4 long-horizon extrinsic manipulation tasks, each with up to 4 primitives. Our experiments cover 10 objects and 6 environment configurations. We further show empirically that our method admits a wide range of demonstrations, and that contact retargeting is indeed the key to successfully combining primitives for long-horizon extrinsic manipulation. Code and additional details are available at stanford-tml.github. io/extrinsic-manipulation.

We address the important problem of generalizing robotic rearrangement to clutter without any explicit object models. We first generate over 650K cluttered scenes-orders of magnitude more than prior work-in diverse everyday environments, such as cabinets and shelves. We render synthetic partial point clouds from this data and use it to train our CabiNet model architecture. CabiNet is a collision model that accepts object and scene point clouds, captured from a single-view depth observation, and predicts collisions for SE(3) object poses in the scene. Our representation has a fast inference speed of 7μs/query with nearly 20% higher performance than baseline approaches in challenging environments. We use this collision model in conjunction with a Model Predictive Path Integral (MPPI) planner to generate collision-free trajectories for picking and placing in clutter. CabiNet also predicts waypoints, computed from the scene's signed distance field (SDF), that allows the robot to navigate tight spaces during rearrangement. This improves rearrangement performance by nearly 35% compared to baselines. We systematically evaluate our approach, procedurally generate simulated experiments, and demonstrate that our approach directly transfers to the real world, despite training exclusively in simulation. Supplementary material and videos of robot experiments in completely unknown scenes are available at: cabinet-object-rearrangement.github.io.

Grasping is the process of picking up an object by applying forces and torques at a set of contacts. Recent advances in deep learning methods have allowed rapid progress in robotic object grasping. In this systematic review, we surveyed the publications over the last decade, with a particular interest in grasping an object using all six degrees of freedom of the end-effector pose. Our review found four common methodologies for robotic grasping: sampling-based approaches, direct regression, reinforcement learning, and exemplar approaches In addition, we found two “supporting methods” around grasping that use deep learning to support the grasping process, shape approximation, and affordances. We have distilled the publications found in this systematic review (85 papers) into ten key takeaways we consider crucial for future robotic grasping and manipulation research.

Robotic object rearrangement combines the skills of picking and placing objects. When object models are unavailable, typical collision-checking models may be unable to predict collisions in partial point clouds with occlusions, making generation of collision-free grasping or placement trajectories challenging. We propose a learned collision model that accepts scene and query object point clouds and predicts collisions for 6DOF object poses within the scene. We train the model on a synthetic set of 1 million scene/object point cloud pairs and 2 billion collision queries. We leverage the learned collision model as part of a model predictive path integral (MPPI) policy in a tabletop rearrangement task and show that the policy can plan collision-free grasps and placements for objects unseen in training in both simulated and physical cluttered scenes with a Franka Panda robot. The learned model outperforms both traditional pipelines and learned ablations by 9.8% in accuracy on a dataset of simulated collision queries and is 75x faster than the best-performing baseline. Videos and supplementary material are available at https://research.nvidia.com/publication/2021-03_Object-Rearrangement-Using.

We introduce ACRONYM, a dataset for robot grasp planning based on physics simulation. The dataset contains 17.7M parallel-jaw grasps, spanning 8872 objects from 262 different categories, each labeled with the grasp result obtained from a physics simulator. We show the value of this large and diverse dataset by using it to train two state-of-the-art learning-based grasp planning algorithms. Grasp performance improves significantly when compared to the original smaller dataset. Data and tools can be accessed at https://sites.google.com/nvidia.com/graspdataset.

In many applications, including logistics and manufacturing, robot manipulators operate in semi-structured environments alongside humans or other robots. These environments are largely static, but they may contain some movable obstacles that the robot must avoid. Manipulation tasks in these applications are often highly repetitive, but require fast and reliable motion planning capabilities, often under strict time constraints. Existing preprocessing-based approaches are beneficial when the environments are highly-structured, but their performance degrades in the presence of movable obstacles, since these are not modelled a priori. We propose a novel preprocessing-based method called Alternative Paths Planner (APP) that provides provably fixed-time planning guarantees in semi-structured environments. APP plans a set of alternative paths offline such that, for any configuration of the movable obstacles, at least one of the paths from this set is collision-free. During online execution, a collision-free path can be looked up efficiently within a few microseconds. We evaluate APP on a 7 DoF robot arm in semi-structured domains of varying complexity and demonstrate that APP is several orders of magnitude faster than state-of-the-art motion planners for each domain. We further validate this approach with real-time experiments on a robotic manipulator.

To teach robots skills, it is crucial to obtain data with supervision. Since annotating real world data is time-consuming and expensive, enabling robots to learn in a self- supervised way is important. In this work, we introduce a robot system for self-supervised 6D object pose estimation. Starting from modules trained in simulation, our system is able to label real world images with accurate 6D object poses for self-supervised learning. In addition, the robot interacts with objects in the environment to change the object configuration by grasping or pushing objects. In this way, our system is able to continuously collect data and improve its pose estimation modules. We show that the self-supervised learning improves object segmentation and 6D pose estimation performance, and consequently enables the system to grasp objects more reliably. A video showing the experiments can be found at https://youtu.be/W1Y0Mmh1Gd8.

Grasping in cluttered environments is a fundamental but challenging robotic skill. It requires both reasoning about unseen object parts and potential collisions with the manipulator. Most existing data-driven approaches avoid this problem by limiting themselves to top-down planar grasps which is insufficient for many real-world scenarios and greatly limits possible grasps. We present a method that plans 6-DOF grasps for any desired object in a cluttered scene from partial point cloud observations. Our method achieves a grasp success of 80.3%, outperforming baseline approaches by 17.6% and clearing 9 cluttered table scenes (which contain 23 unknown objects and 51 picks in total) on a real robotic platform. By using our learned collision checking module, we can even reason about effective grasp sequences to retrieve objects that are not immediately accessible. Supplementary video can be found here.

Generating grasp poses is a crucial component for any robot object manipulation task. In this work, we formulate the problem of grasp generation as sampling a set of grasps using a variational autoencoder and assess and refine the sampled grasps using a grasp evaluator model. Both Grasp Sampler and Grasp Refinement networks take 3D point clouds observed by a depth camera as input. We evaluate our approach in simulation and real-world robot experiments. Our approach achieves 88% success rate on various commonly used objects with diverse appearances, scales, and weights. Our model is trained purely in simulation and works in the real-world without any extra steps.

It is difficult to create robust, reusable, and reactive behaviors for robots that can be easily extended and combined. Frameworks such as Behavior Trees are flexible but difficult to characterize, especially when designing reactions and recovery behaviors to consistently converge to a desired goal condition. We propose a framework which we call Robust Logical-Dynamical Systems (RLDS), which combines the advantages of task representations like behavior trees with theoretical guarantees on performance. RLDS can also be constructed automatically from simple sequential task plans and will still achieve robust, reactive behavior in dynamic real-world environments. In this work, we describe both our proposed framework and a case study on a simple household manipulation task, with examples for how specific pieces can be implemented to achieve robust behavior. Finally, we show how in the context of these manipulation tasks, a combination of an RLDS with planning can achieve better results under adversarial conditions.

Interactive perception exploits the correlation between forceful interactions and changes in the observed signals to extract task-relevant information from the sensor stream. Finding the most informative interactions to perceive complex objects, like articulated mechanisms, is challenging because the outcome of the interaction is difficult to predict. We propose a method to select the most informative action while deriving a model of articulated mechanisms that includes kinematic, geometric, and dynamic properties. Our method addresses the complexity of the action selection task based on two insights. First, we show that for a class of interactive perception methods, information gain can be approximated by the amount of motion induced in the mechanism. Second, we resort to physics simulations grounded in the real-world through interactive perception to predict possible action outcomes. Our method enables the robot to autonomously select actions for interactive perception that reveal most information, given the current knowledge of the world. This leads to improved perception and more accurate world models, finally enabling robust manipulation.

We describe the winning entry to the Amazon Picking Challenge 2015. From the experience of building this system and competing, we derive several conclusions: (1) We suggest to characterize robotic system building along four key aspects, each of them spanning a spectrum of solutions - modularity vs. integration, generality vs. assumptions, computation vs. embodiment, and planning vs. feedback. (2) To understand which region of each spectrum most adequately addresses which robotic problem, we must explore the full spectrum of possible approaches. (3) For manipulation problems in unstructured environments, certain regions of each spectrum match the problem most adequately, and should be exploited further. This is supported by the fact that our solution deviated from the majority of the other challenge entries along each of the spectra. This is an abridged version of a conference publication.

Environment-constrained grasping exploits beneficial interactions between hand, object, and environment to increase grasp success. Instead of focusing on the final static relationship between hand posture and object pose, this view of grasping emphasizes the need and the opportunity to select the most appropriate, contact-rich grasping motion, leading up to a final static grasp configuration. This view changes the nature of the underlying planning problem: Instead of planning for static contact points, we need to decide which environmental constraint (EC) to use during the grasping motion. We propose a method to make these decisions based on depth measurements so as to generate robust grasps for a large variety of objects. Our planner exploits the advantages of a soft robot hand and learns a hand-specific classifier for edge-, surface-, and wall-grasps, each exploiting a different EC. Additionally, we show how the model can continuously be improved in a contextual multi-armed bandit setting without an explicit training and test phase, enabling the continuous improvement of a robot's grasping skills throughout life time.

In this paper we present a planner that interleaves free-space motion with motion in contact to reduce uncertainty. The planner finds such motions by growing a search tree in the combined space of collision-free and contact configurations. The planner reasons efficiently about the accumulated uncertainty by factoring the state in a belief over configuration and a fully observable contact state. We show the uncertainty-reducing capabilities of the planner on manipulation benchmark from the POMDP literature. The planner scales up to more complex problems like manipulation under uncertainty in seven-dimensional configuration space. We validate our planner in simulation and on a real robot.

We present an automatic method for interactive control of physical humanoid robots based on high-level tasks that does not require manual specification of motion trajectories or specially-designed control policies. The method is based on the combination of a model-based policy that is trained off-line in simulation and sends high-level commands to a model-free controller that executes these commands on the physical robot. This low-level controller simultaneously learns and adapts a local model of dynamics on-line and computes optimal controls under the learned model. The high-level policy is trained using a combination of trajectory optimization and neural network learning, while considering physical limitations such as limited sensors and communication delays. The entire system runs in real-time on the robot's computer and uses only on-board sensors. We demonstrate successful policy execution on a range of tasks such as leaning, hand reaching, and robust balancing behaviors atop a tilting base on the physical robot and in simulation.

Dexterous multi-fingered hands can accomplish fine manipulation behaviors that are infeasible with simple robotic grippers. However, sophisticated multi-fingered hands are often expensive and fragile. Low-cost soft hands offer an appealing alternative to more conventional devices, but present considerable challenges in sensing and actuation, making them difficult to apply to more complex manipulation tasks. In this paper, we describe an approach to learning from demonstration that can be used to train soft robotic hands to perform dexterous manipulation tasks. Our method uses object-centric demonstrations, where a human demonstrates the desired motion of manipulated objects with their own hands, and the robot autonomously learns to imitate these demonstrations using reinforcement learning. We propose a novel algorithm that allows us to blend and select a subset of the most feasible demonstrations, which we use with an extension of the guided policy search framework that learns generalizable neural network policies. We demonstrate our approach on the RBO Hand 2, with learned motor skills for turning a valve, manipulating an abacus, and grasping.

We present a method for multi-class segmentation from RGB-D data in a realistic warehouse picking setting. The method computes pixel-wise probabilities and combines them to find a coherent object segmentation. It reliably segments objects in cluttered scenarios, even when objects are translucent, reflective, highly deformable, have fuzzy surfaces, or consist of loosely coupled components. The robust performance results from the exploitation of problem structure inherent to the warehouse setting. The proposed method proved its capabilities as part of our winning entry to the 2015 Amazon Picking Challenge. We present a detailed experimental analysis of the contribution of different information sources, compare our method to standard segmentation techniques, and assess possible extensions that further enhance the algorithm's capabilities. We release our software and data sets as open source.

There is strong evidence that robustness in human and robotic grasping can be achieved through the deliberate exploitation of contact with the environment. In contrast to this, traditional grasp planners generally disregard the opportunity to interact with the environment during grasping. In this paper, we propose a novel view of grasp planning that centers on the exploitation of environmental contact. In this view, grasps are sequences of constraint exploitations, i.e. consecutive motions constrained by features in the environment, ending in a grasp. To be able to generate such grasp plans, it becomes necessary to consider planning, perception, and control as tightly integrated components. As a result, each of these components can be simplified while still yielding reliable grasping performance. We propose a first implementation of a grasp planner based on this view and demonstrate in real-world experiments the robustness and versatility of the resulting grasp plans.

As a first step towards transferring human grasping capabilities to robots, we analyzed the grasping behavior of human subjects. We derived a taxonomy in order to adequately represent the observed strategies. During the analysis of the recorded data, this classification scheme helped us to obtain a better understanding of human grasping behavior. We will provide support for our hypothesis that humans exploit compliant contact between the hand and the environment to compensate for uncertainty. We will also show a realization of the resulting grasping strategies on a real robot. It is our belief that the detailed analysis of human grasping behavior will ultimately lead to significant increases in robot manipulation and dexterity.

In grasping, shape adaptation between hand and object has a major influence on grasp success. In this paper, we present an approach to grasping unknown objects that explicitly considers the effect of shape adaptability to simplify perception. Shape adaptation also occurs between the hand and the environment, for example, when fingers slide across the surface of the table to pick up a small object. Our approach to grasping also considers environmental shape adaptability to select grasps with high probability of success. We validate the proposed shape-adaptability-aware grasping approach in 880 real-world grasping trials with 30 objects. Our experiments show that the explicit consideration of shape adaptability of the hand leads to robust grasping of unknown objects. Simple perception suffices to achieve this robust grasping behavior.

In this paper, we present an approach that allows a robot to observe, generalize, and reproduce tasks observed from multiple demonstrations. Motion capture data is recorded in which a human instructor manipulates a set of objects. In our approach, we learn relations between body parts of the demonstrator and objects in the scene. These relations result in a generalized task description. The problem of learning and reproducing human actions is formulated using a dynamic Bayesian network (DBN). The posteriors corresponding to the nodes of the DBN are estimated by observing objects in the scene and body parts of the demonstrator. To reproduce a task, we seek for the maximum-likelihood action sequence according to the DBN. We additionally show how further constraints can be incorporated online, for example, to robustly deal with unforeseen obstacles. Experiments carried out with a real 6-DoF robotic manipulator as well as in simulation show that our approach enables a robot to reproduce a task carried out by a human demonstrator. Our approach yields a high degree of generalization illustrated by performing a pick-and-place and a whiteboard cleaning task.

Wheeled tour guide robots have already been deployed in various museums or fairs worldwide. A key requirement for successful tour guide robots is to interact with people and to entertain them. Most of the previous tour guide robots, however, focused more on the involved navigation task than on natural interaction with humans. Humanoid robots, on the other hand, offer a great potential for investigating intuitive, multimodal interaction between humans and machines. In this paper, we present our mobile full-body humanoid tour guide robot Robotinho. We provide mechanical and electrical details and cover perception, the integration of multiple modalities for interaction, navigation control, and system integration aspects. The multimodal interaction capabilities of Robotinho have been designed and enhanced according to the questionnaires filled out by the people who interacted with the robot at previous public demonstrations. We present experiences we have made during experiments in which untrained users interacted with the robot.