A grasp generation technique for robotic pick-up of parts. A database of solid or surface models is provided for all objects and grippers which are to be evaluated. A gripper is selected and a random initialization is performed, where random objects and poses are selected from the object database. An iterative optimization computation is then performed, where many hundreds of grasps are computed for each part with surface contact between the part and the gripper, and sampling for grasp diversity and global optimization. Finally, a physical environment simulation is performed, where the grasps for each part are mapped to simulated piles of objects in a bin scenario. The grasp points and approach directions from the physical environment simulation are then used to train neural networks for grasp learning in real-world robotic operations, where the simulation results are correlated to camera depth image data to identify a high quality grasp.

A method and arrangement for testing and/or correction of a weld (34, 36, 38) of a test object (26, 102), including alignment of an ultrasonic probe (16, 128) guided by a robot (100) on a target position of the weld (28, 30, 32), determination of the actual position (34, 36, 38) of the weld by means of an optical sensor (22, 130) and alignment of the ultrasonic probe (16) on the actual position, and measurement of the weld, where CAD data of the target position of the weld (28, 30, 32) is made available, on the basis of the CAD data of the weld the ultrasonic probe (16, 128) is aligned on the target position of the weld, and the ultrasonic probe is placed on the weld with controlled force after determination of the actual position (34, 36, 38) of the weld by means of the optical sensor (22, 130).

A method for generating a training data set for training an industrial robot which can be trained based on a corresponding training data set, comprising: providing a first imaging information, which describes a first one- or multi-dimensional image of an object which is to be relocated by means of an industrial robot which is to be trained on the basis of the training data set to be generated; processing the first imaging information to generate further imaging information, which describes at least one artificially generated further one- or multi-dimensional image of the object which is to be moved by means of an industrial robot which is to be trained on the basis of the training data set to be generated; and processing the further imaging information to generate a training data set for training an industrial robot which can be trained on the basis of the training data set.

A technique for displaying a representative path associated with a robotic device. The technique includes detecting at least one reference point within a first image of a workspace, generating the representative path based on path instructions associated with the robotic device and the at least one reference point, and displaying the representative path within the workspace.

Embodiments of the present invention provide methods and systems to analyze wearable technology. A robot with snake assembly works in conjunction with a server in order to simulate the locomotive actions of appendages and to concomitantly determine the response of wearable technology devices, which are attached to the snake robot assembly, to the simulated locomotive actions.

A robot system includes: a recognition means that recognizes an operator in distance image data as a security surveillance target; an area setting means that sets, in the distance image data, a common work area that a robot and the operator are allowed to enter; and a boundary setting means that sets, in the distance image data, a first boundary that the operator S is allowed to cross for entering the common work area but the robot is not allowed to cross, and a second boundary that the robot is allowed to cross for entering the common work area but the operator is not allowed to cross, wherein the crossing of the second boundary by the security surveillance target and the crossing of the first boundary by a non-target object not recognized as the security surveillance target are detected.

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.

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.

One embodiment of the present invention sets forth a technique for generating simulated training data for a physical process. The technique includes receiving, as input to at least one machine learning model, a first simulated image of a first object, wherein the at least one machine learning model includes mappings between simulated images generated from models of physical objects and real-world images of the physical objects. The technique also includes performing, by the at least one machine learning model, one or more operations on the first simulated image to generate a first augmented image of the first object. The technique further includes transmitting the first augmented image to a training pipeline for an additional machine learning model that controls a behavior of the physical process.

Example implementations relate to robotic operations libraries. An example library may include sets of operation instructions and other information for robotic devices to use to complete desired tasks. For instance, a respective set of operation instructions is determined based on successive simulations in which a virtual robotic device comprising an adjustable configuration initially based on the given configuration of a robotic device performs operations related to a task in an adjustable virtual environment until one or more simulations result in the virtual robotic device performing respective operations that complete the task at a success level that satisfies a predefined threshold. The library may provide a set of instructions for performing operations to a robotic device based on a query received from the robotic device that includes information indicative of a configuration and an environment of the robotic device.