Abstract: One embodiment of a method for generating an articulation model includes receiving a first set of images of an object in a first articulation and a second set of images of the object in a second articulation, performing one or more operations to generate first three-dimensional (3D) geometry based on the first set of images, performing one or more operations to generate second 3D geometry based on the second set of images, and performing one or more operations to generate an articulation model of the object based on the first 3D geometry and the second 3D geometry.

Abstract: A machine-learning control system is trained to perform a task using a simulation. The simulation is governed by parameters that, in various embodiments, are not precisely known. In an embodiment, the parameters are specified with an initial value and expected range. After training on the simulation, the machine-learning control system attempts to perform the task in the real world. In an embodiment, the results of the attempt are compared to the expected results of the simulation, and the parameters that govern the simulation are adjusted so that the simulated result matches the real-world attempt. In an embodiment, the machine-learning control system is retrained on the updated simulation. In an embodiment, as additional real-world attempts are made, the simulation parameters are refined and the control system is retrained until the simulation is accurate and the control system is able to successfully perform the task in the real world.

Abstract: Apparatuses, systems, and techniques to determine a grasp and placement of an object in an environment. In at least one embodiment, one or more neural networks are used to identify one or more grasping and placement masks used to manipulate an object.

Abstract: Apparatuses, systems, and techniques to generate data to train a robotic device to perform tasks. In at least one embodiment, one or more first videos of a robotic device performing a task is used to generate one or more second videos of the robotic device performing the task differently than depicted in the one or more first videos.

Abstract: In various examples, systems and methods are disclosed relating to training machine learning models using human demonstration of segments of a task, where other segments of the task are performed by a planning method, such as a Task and Motion Planning (TAMP) system. A method may include segmenting a task to be performed by a robot into segments, determining a first set of instructions of a plurality of sets of instructions for operating the robot to perform a first objective of a first segment, determining that the plurality of sets of instructions is inadequate to perform a second objective of a second segment, receiving from a user device a second set of instructions for operating the robot for the second segment following an end of the first segment, and updating a machine learning model for controlling the robot using the second set of instructions for the second segment.

Abstract: A robotic system is provided for performing rearrangement tasks guided by a natural language instruction. The system can include a number of neural networks used to determine a selected rearrangement of the objects in accordance with the natural language instruction. A target object predictor network processes a point cloud of the scene and the natural language instruction to identify a set of query objects that are to-be-rearranged. A language conditioned prior network processes the point cloud, natural language instruction, and the set of query objects to sample a distribution of rearrangements to generate a number of sets of pose offsets for the set of query objects. A discriminator network then processes the samples to generate scores for the samples. The samples may be refined until a score for at least one of the sample generated by the discriminator network is above a threshold value.

Abstract: A technique for training a neural network, including generating a plurality of input vectors based on a first plurality of task demonstrations associated with a first robot performing a first task in a simulated environment, wherein each input vector included in the plurality of input vectors specifies a sequence of poses of an end-effector of the first robot, and training the neural network to generate a plurality of output vectors based on the plurality of input vectors. Another technique for generating a task demonstration, including generating a simulated environment that includes a robot and at least one object, causing the robot to at least partially perform a task associated with the at least one object within the simulated environment based on a first output vector generated by a trained neural network, and recording demonstration data of the robot at least partially performing the task within the simulated environment.

Abstract: Apparatuses, systems, and techniques to identify at least one physical characteristic of materials from computer simulations of manipulations of materials. In at least one embodiment, physical characteristics are determined by comparing measured statistics of observed manipulations to simulations of manipulations using a simulator trained with a likelihood-free inference engine.

Abstract: In various examples, 3D object knowledge can be developed to extract diverse knowledge from large language models, and a part-grounding model can be trained to ground part semantics in terms of local shape features and spatial relations between parts. For example, knowledge that “the opening part of a mug that affords the pouring action is located on the top of the mug body and is often circular” can be grounded by identifying a previously unknown “opening” part based on its spatial relation to the known “body” part and its circular shape. A robotic system, for example, may use a model to identify an unlabeled part of a 3D object in imaging data. The model may be generated using natural language descriptions of relationships between parts of 3D objects, with descriptions generated using a language model that produces text in response to queries related to spatial relationships between the parts.

Abstract: Apparatuses, systems, and techniques to grasp objects with a robot. In at least one embodiment, a neural network is trained to determine a grasp pose of an object within a cluttered scene using a point cloud generated by a depth camera.

Abstract: In various examples, an autonomous system may use a multi-stage process to solve three-dimensional (3D) manipulation tasks from a minimal number of demonstrations and predict key-frame poses with higher precision. In a first stage of the process, for example, the disclosed systems and methods may predict an area of interest in an environment using a virtual environment. The area of interest may correspond to a predicted location of an object in the environment, such as an object that an autonomous machine is instructed to manipulate. In a second stage, the systems may magnify the area of interest and render images of the virtual environment using a 3D representation of the environment that magnifies the area of interest. The systems may then use the rendered images to make predictions related to key-frame poses associated with a future (e.g., next) state of the autonomous machine.

Abstract: Apparatuses, systems, and techniques to identify at least one physical characteristic of materials from computer simulations of manipulations of materials. In at least one embodiment, physical characteristics are determined by comparing measured statistics of observed manipulations to simulations of manipulations using a simulator trained with a likelihood-free inference engine.

Abstract: One embodiment of a method for generating simulation data to train a machine learning model includes generating a plurality of simulation environments based on a user input, and for each simulation environment included in the plurality of simulation environments: generating a plurality of tasks for a robot to perform within the simulation environment, performing one or more operations to determine a plurality of robot trajectories for performing the plurality of tasks, and generating simulation data for training a machine learning model by performing one or more operations to simulate the robot moving within the simulation environment according to the plurality of trajectories.

Abstract: A robot is controlled using a combination of model-based and model-free control methods. In some examples, the model-based method uses a physical model of the environment around the robot to guide the robot. The physical model is oriented using a perception system such as a camera. Characteristics of the perception system may be are used to determine an uncertainty for the model. Based at least in part on this uncertainty, the system transitions from the model-based method to a model-free method where, in some embodiments, information provided directly from the perception system is used to direct the robot without reliance on the physical model.

Abstract: One embodiment of a method for controlling a robot includes receiving sensor data indicating a state of the robot, generating an action based on the sensor data and a trained machine learning model, computing a target state of the robot based on the action and a previous target state of the robot, and causing the robot to move based on the target state of the robot.

Abstract: One embodiment of a method for training a machine learning model to control a robot includes causing a model of the robot to move within a simulation based on one or more outputs of the machine learning model, computing an error within the simulation, computing at least one of a reward or an observation based on the error, and updating one or more parameters of the machine learning model based on the at least one of a reward or an observation.

Abstract: In various examples, a machine may generate, using sensor data capturing one or more views of an environment, a virtual environment including a 3D representation of the environment. The machine may render, using one or more virtual sensors in the virtual environment, one or more images of the 3D representation of the environment. The machine may apply the one or more images to one or more machine learning models (MLMs) trained to generate one or more predictions corresponding to the environment. The machine may perform one or more control operations based at least on the one or more predictions generated using the one or more MLMs.

Abstract: Apparatuses, systems, and techniques to generate control commands. In at least one embodiment, control commands are generated based on, for example, one or more images depicting a hand.

Abstract: One common robotic task is the rearrangement of physical objects situated in an environment. This typically involves a robot manipulator picking up a target object and placing the target object in some target location, such as a shelf, cabinet or cubby, and requires the skills of picking, placing and generating complex collision-free motions, oftentimes in a cluttered environment. The present disclosure provides collision detection for object rearrangement using a three-dimensional (3D) scene representation.

Abstract: Apparatuses, systems, and techniques to control a real-world and/or virtual device (e.g., a robot). In at least one embodiment, the device is controlled based, at least in part on, for example, one or more neural networks. Parameter values for the neural network(s) may be obtained by training the neural network(s) to control movement of a first agent with respect to at least one first target while avoiding collision with at least one stationary first holder of the at least one first target, and updating the parameter values by training the neural network(s) to control movement of a second agent with respect to at least one second target while avoiding collision with at least one non-stationary second holder of the at least one second target.