Methods, systems, and apparatus, including computer programs encoded on computer storage media for evaluating robot learning. In some implementations, a system receives classification examples from a plurality of remote devices over a communication network. The classification examples can include (i) a data representation generated by a remote device based on sensor data captured by the remote device and (ii) a classification corresponding to the data representation. The system assigns quality scores to the classification examples based on a level of similarity of the data representations with other data representations. The system selects a subset of the classification examples based on the quality scores assigned to the classification examples. The system trains a machine learning model using the selected subset of the classification examples.

Methods, systems, and apparatus, including computer-readable media storing executable instructions, for enhancing robot learning. In some implementations, a robot stores first embeddings generated using a first machine learning model, and the first embeddings include one or more first private embeddings that are not shared with other robots. The robot receives a second machine learning model from a server system over a communication network. The robot generates a second private embedding for each of the one or more first private embeddings using the second machine learning model. The robot adds the second private embeddings to the cache of the robot and removes the one or more first private embeddings from the cache of the robot.

Remote control robot system includes a master device, slave arm having plurality of control modes including automatic and manual mode, control device configured to operate slave arm, an entering-person sensing device configured to detect entering person into operational area of slave arm, entering-person identifying information acquisition device configured to acquire entering-person identifying information for identifying whether entering person is operator who carries master device, and operation regulating device configured to regulate operation of slave arm based on information acquired from entering-person sensing device and information acquisition device. In automatic mode, operation regulating device regulates operation of slave arm when entering person is detected. In manual mode, operation regulating device allows operation of slave arm to continue when entering person is detected and entering person is operator, and regulates operation of the slave arm when entering person is other than operator.

A teaching device includes an image acquisition unit which acquires an image including a manipulation target object which is linked to operations of an arm of a robot and a teaching position of the manipulation target object, a movement control unit which controls the arm to move the manipulation target object in the image to the teaching position, and a teaching information acquisition unit which acquires a state of the arm in a state in which the manipulation target object in the image is present at the teaching position as teaching information.

In one embodiment, a method includes receiving, from a first sensor on a robot, first sensor data indicative of an environment of the robot. The method also includes identifying, based on the first sensor data, an object of an object type in the environment of the robot, where the object type is associated with a classifier that takes sensor data from a predetermined pose relative to the object as input. The method further includes causing the robot to position a second sensor on the robot at the predetermined pose relative to the object. The method additionally includes receiving, from the second sensor, second sensor data indicative of the object while the second sensor is positioned at the predetermined pose relative to the object. The method further includes determining, by inputting the second sensor data into the classifier, a property of the object.

Methods, systems, and apparatus, including computer programs encoded on computer storage media for evaluating robot learning. In some implementations, one or more computers receive object classification examples from a plurality of robots. Each object classification example includes (i) an embedding that a robot generated using a machine learning model, and (ii) an object classification corresponding to the embedding. The object classification examples are evaluated based on a similarity of the received embeddings with respect to other embeddings. A subset of the object classification examples is selected based on the evaluation of the quality of the embeddings. The subset of the object classification examples is distributed to the robots in the plurality of robots.

A remote control robot system includes a slave arm, a master main body imitating the shape of an object handled by the slave arm, a manipulation receiving device configured to receive manipulation of an operator based on the position and posture of the master main body, and a control device configured to control operation of the slave arm based on the manipulation received by the manipulation receiving device so that behavior of the object corresponds to behavior of the master main body.

A remote-control manipulator system includes a manipulator configured to receive a manipulating instruction from an operator, a slave arm configured to perform a series of works comprised of a plurality of processes, a camera configured to image operation of the slave arm, a display device configured to display an image captured by the camera, a storage device configured to store information related to environment in a workspace as an environment model, and a control device. The control device is configured, while operating the slave arm manually or hybridly, to acquire circumference information that is information related to a circumference area of an area imaged by the camera based on the environment model stored in the storage device, and display on the display device so that the image captured by the camera and the circumference information are interlocked.

A control device includes a learning control part in which a difference is calculated between a target position and an actual position of a portion detected based on a sensor, and an operation-speed change rate is increased or reduced several times within a maximum value of the operation-speed change rate set for increasing or reducing the operation speed of a robot mechanism unit and within allowance conditions of vibrations occurring at the portion to be controlled; meanwhile, learning is repeated to calculate an updated compensation amount based on the difference and a previous compensation amount previously calculated for suppressing vibrations at each operation-speed change rate, and a convergent compensation amount and a convergent operation-speed change rate are stored after convergence of the compensation amount and the operation-speed change rate.

In one embodiment, a method includes receiving, from a first sensor on a robot, first sensor data indicative of an environment of the robot. The method also includes identifying, based on the first sensor data, an object of an object type in the environment of the robot, where the object type is associated with a classifier that takes sensor data from a predetermined pose relative to the object as input. The method further includes causing the robot to position a second sensor on the robot at the predetermined pose relative to the object. The method additionally includes receiving, from the second sensor, second sensor data indicative of the object while the second sensor is positioned at the predetermined pose relative to the object. The method further includes determining, by inputting the second sensor data into the classifier, a property of the object.