A technique for automatically finding a collision-free return-to-home path for a robot. The technique includes running a simulated virtual 3D environment which emulates the physical robot and workcell in real time, including the positions and poses of all robots, workpieces and obstacles in the workcell. Upon request by an operator, a return-to-home path search is executed based on the virtual 3D environment, where the path search calculates a solution which moves the robot from a current position to its home or recovery position while avoiding collisions with other robots, workpieces or objects in the workcell. In addition to collision avoidance, the path search considers other constraints such as prohibited zones in the workspace and robot joint positions. When the recovery path is computed, the solution program is sent back to the physical environment for execution by the physical robot.

A computer-implemented method including determining a group to which a first object belongs and a group to which a second object belongs, executing a simulation including the first object and the second object, executing a collision determination between the first object and the second object during execution of the simulation, and changing the group to which the first object belongs when a predetermined condition is satisfied. The collision determination is executed only when the group to which the first object belongs is different from the group to which the second object belongs.

Various embodiments include prediction model learning methods for industrial systems, including a simulation according to a simulation task on a platform. Some methods include: using statistical metrics to quantify a simulation task to extract features; extracting parameter groups from system modules, adjusting the values of the parameter groups, and triggering a simulation for the industrial system on the simulation platform based on a plurality of parameter groups having different values; recording performance metrics according to the values of the parameter groups after the parameter adjustment; and training data with a machine learning algorithm based on the features of the simulation task, the corresponding parameter groups having different values and the performance metrics, and generating a prediction model.

One embodiment of the present invention sets forth a technique for controlling the execution of a physical process. The technique includes receiving, as input to a machine learning model that is configured to adapt a simulation of the physical process executing in a virtual environment to a physical world, simulated output for controlling how the physical process performs a task in the virtual environment and real-world data collected from the physical process performing the task in the physical world. The technique also includes performing, by the machine learning model, one or more operations on the simulated output and the real-world data to generate augmented output. The technique further includes transmitting the augmented output to the physical process to control how the physical process performs the task in the physical world.

Systems and a method predict a generation of a robotic program for industrial coating. Inputs are received including a virtual representation of a robot, a coating gun, elements of the object surface to be coated and a set of desired coating thickness ranges. Inputs on a coating dispersion object are also received. Training data of a plurality of robotic programs for industrial coating and of their corresponding coating thickness coverage on a plurality of surfaces are received. The training data are processed in x, y tuples so as to learn a mapping function to generate a coating prediction module. Starting with a given selected valid thickness coverage as input parameters, it is proceeded in an iterative manner to predict a robotic program via the coating prediction module. A coating robotic program is generated for each surface element based on the resulting predicted coating programs.

An acquisition unit acquires operation data for expressing real-time motions of a plurality of robotic devices. A virtual robot control unit uses the operation data regarding the plurality robotic devices to move a plurality of virtual robots corresponding to the plurality of robotic devices in the same virtual space. An image generating unit generates an image of the virtual space in which the plurality of virtual robots are in motion. The virtual robot control unit makes the plurality of virtual robots compete in a virtual sports venue.

Methods, apparatus, and computer-readable media for determining and utilizing corrections to robot actions. Some implementations are directed to updating a local features model of a robot in response to determining a human correction of an action performed by the robot. The local features model is used to determine, based on an embedding generated over a corresponding neural network model, one or more features that are most similar to the generated embedding. Updating the local features model in response to a human correction can include updating a feature embedding, of the local features model, that corresponds to the human correction. Adjustment(s) to the features model can immediately improve robot performance without necessitating retraining of the corresponding neural network model.

A processing path generating device including an intuitive path teaching device and a controller is provided. The intuitive path teaching device is provided for gripping and moving with respect to a workpiece to create a moving path. The intuitive path teaching device has a detecting portion for detecting a surface feature of the workpiece. The controller is connected to the intuitive path teaching device. The controller generates a processing path according to the moving path of the intuitive path teaching device and the surface feature of the workpiece.

An autonomous mobile robot includes a first arithmetic unit configured to calculate a course direction based on an own position, a moving-object position, and a moving-object velocity vector, the course direction being a direction in which the autonomous mobile robot should travel, a second arithmetic unit configured to input the own position, the moving-object position, the moving-object velocity vector, and the course direction into a trained model and thereby calculate an estimated position, the trained model being a model that has been trained, the estimated position being a position at which the autonomous mobile robot is estimated to arrive a predetermined time later without colliding with the moving object, a generating unit configured to generate a remaining route from the estimated position to a destination, and a movement control unit configured to control a movement to the destination based on the course direction and the remaining route.

In various examples, a computer-implemented method of generating instructions for a welding robot. The computer-implemented method comprises identifying an expected position of a candidate seam on a part to be welded based on a Computer Aided Design (CAD) model of the part, scanning a workspace containing the part to produce a representation of the part, identifying the candidate seam on the part based on the representation of the part and the expected position of the candidate seam, determining an actual position of the candidate seam, and generating welding instructions for the welding robot based at least in part on the actual position of the candidate seam.