Abstract: In various examples, systems and methods are disclosed relating to geometric fabrics for accelerated policy learning and sim-to-real transfer in robotics systems, platforms, and/or applications. For example, a system can provide an input indicative of a goal pose for a robot to a model to cause the model to generate an output, the output representing a plurality of points along a path for movement of the robot to the goal pose; and generate one or more control signals for operation of the robot based at least on the plurality of points along the path and a policy corresponding to one or more criteria for the operation of the robot. In examples, the system can provide the one or more control signals to the robot to cause the robot to move toward the goal pose.

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: A decoupling control method for a humanoid robot includes: decomposing tasks of the humanoid robot to obtain kinematic tasks and dynamic tasks, and classifying corresponding joints of the humanoid robot into kinematic task joints or dynamic task joints; solving desired positions and desired speeds of the kinematic task joints for performing the kinematic tasks according to desired positions and desired speeds of ends in the kinematic tasks using inverse kinematics; calculating torques of the kinematic task joints based on the desired positions and desired speeds of the kinematic task joints; and solving a pre-built optimization model of torques required for the dynamic task joints based on the calculated torques of the kinematic task joints, to obtain torques required by the dynamic task joints for performing the dynamic tasks.

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: A decoupling control method for a humanoid robot includes: decomposing tasks of the humanoid robot to obtain kinematic tasks and dynamic tasks, and classifying corresponding joints of the humanoid robot into kinematic task joints or dynamic task joints; solving desired positions and desired speeds of the kinematic task joints for performing the kinematic tasks according to desired positions and desired speeds of ends in the kinematic tasks using inverse kinematics; calculating torques of the kinematic task joints based on the desired positions and desired speeds of the kinematic task joints; and solving a pre-built optimization model of torques required for the dynamic task joints based on the calculated torques of the kinematic task joints, to obtain torques required by the dynamic task joints for performing the dynamic tasks.