A computer-implemented method includes recording one or more demonstrations of a task performed by a user. Movements of one or more joints of the user are determined from the one or more demonstrations. By a computer processor, a neural network or Gaussian mixture model incorporating one or more contraction analysis constraints is learned, based on the movements of the one or more joints of the user, the one or more contraction analysis constraints representing motion characteristics of the task. A first initial position of a robot is determined. A first trajectory of the robot is determined to perform the task, based at least in part on the neural network or Gaussian mixture model and the first initial position.

Apparatus and methods for training and operating of robotic devices. Robotic controller may comprise a predictor apparatus configured to generate motor control output. The predictor may be operable in accordance with a learning process based on a teaching signal comprising the control output. An adaptive controller block may provide control output that may be combined with the predicted control output. The predictor learning process may be configured to learn the combined control signal. Predictor training may comprise a plurality of trials. During initial trial, the control output may be capable of causing a robot to perform a task. During intermediate trials, individual contributions from the controller block and the predictor may be inadequate for the task. Upon learning, the control knowledge may be transferred to the predictor so as to enable task execution in absence of subsequent inputs from the controller. Control output and/or predictor output may comprise multi-channel signals.

Apparatus and methods for training and operating of robotic devices. Robotic controller may comprise a predictor apparatus configured to generate motor control output. The predictor may be operable in accordance with a learning process based on a teaching signal comprising the control output. An adaptive controller block may provide control output that may be combined with the predicted control output. The predictor learning process may be configured to learn the combined control signal. Predictor training may comprise a plurality of trials. During initial trial, the control output may be capable of causing a robot to perform a task. During intermediate trials, individual contributions from the controller block and the predictor may be inadequate for the task. Upon learning, the control knowledge may be transferred to the predictor so as to enable task execution in absence of subsequent inputs from the controller. Control output and/or predictor output may comprise multi-channel signals.

A method for computing joint torques applied by actuators to perform a control of a movement of a robot arm having several degrees of freedom is provided. The method includes the act of providing, by a trajectory generator, trajectory vectors specifying a desired trajectory of the robot arm for each degree of freedom. The trajectory vectors are mapped to corresponding latent representation vectors that capture inherent properties of the robot arm using basis functions with trained parameters. The latent representation vectors are multiplied with trained core tensors to compute the joint torques for each degree of freedom.

Systems and a method for predicting a motion trajectory of a robot moving between a given pair of robotic locations. Training data of motion trajectories of the robot are received for a plurality of robotic location pairs. The training data are processed so as to obtain x tuples and y tuples for machine learning purposes; wherein the x tuples describe the robotic location pair and the y tuples describe one or more intermediate robotic locations at specific time stamps during the motion of the robot between the locations of the location pair. From the processed data, a function is learned for mapping the x tuples into the y tuples so as to generate a motion prediction module for the robot. For a given robotic location pair, the robotic motion between the given pair is predicted by obtaining the corresponding intermediate locations resulting from the motion prediction module.

A control and learning module for controlling a robotic arm includes at least one learning module including at least one neural network. The at least one neural network is configured to receive and be trained by both state measurements based on measurements of current state and observation measurements based on observation data during an initial learning phase. The at least one learning module is further configured to be re-tuned by updated observation data for improved performance during an operations and secondary learning phase when the robotic arm is in normal operation and after the initial learning phase.

A learning method is performed by a computer. The method includes: inputting a first image to a model, which outputs, from an input image, candidates for a specific region and confidences indicating probabilities of the respective candidates being the specific region, to cause the model to output a plurality of candidates for the specific region and confidences for the respective candidates; calculating a first value for each of candidates whose confidences do not satisfy a certain criterion among the candidates output by the model, the first value increasing as the confidence increases; calculating a second value obtained by weighting the first value such that the second value decreases as the confidence increases; and updating the model such that the second value decreases.

A control device repeats learning of: calculating an allowable condition for speed variations during a processing operation based on an allowable condition for processing error; setting an operating speed change rate used to increase or reduce an operating speed of a robot mechanism unit using a calculated allowable condition for speed variations; and, while increasing or reducing the operating speed change rate over a plurality of repetitions within a range not exceeding a maximum value of the operating speed change rate and within a range of an allowable condition for vibrations occurring in a control target, calculates a new correction amount based on an amount of difference between a position of the control target detected based on a sensor and a target position, and a previously-calculated correction amount.

Apparatus and methods for training and controlling of, for instance, robotic devices. In one implementation, a robot may be trained by a user using supervised learning. The user may be unable to control all degrees of freedom of the robot simultaneously. The user may interface to the robot via a control apparatus configured to select and operate a subset of the robot's complement of actuators. The robot may comprise an adaptive controller comprising a neuron network. The adaptive controller may be configured to generate actuator control commands based on the user input and output of the learning process. Training of the adaptive controller may comprise partial set training. The user may train the adaptive controller to operate first actuator subset. Subsequent to learning to operate the first subset, the adaptive controller may be trained to operate another subset of degrees of freedom based on user input via the control apparatus.

The automatic operation system includes a plurality of learned imitation models and a model selecting unit. The learned imitation models are constructed by machine learning of operation history data, the operation history data being classified into several groups by an automatic classification system algorithm, the operation history data of each group being learned by the imitation model corresponding to the group. The operation history data include data indicating a surrounding environment and data indicating an operation of an operator in the surrounding environment. The model selecting unit selects one imitation model from several imitation models based on a result of classifying data indicating a given surrounding environment by the automatic classification algorithm of the classification system. The automatic operation system inputs data indicating the surrounding environment to the imitation model selected by the model selecting unit to predict an operation of the operator with respect to the surrounding environment.