Provided is a model parameter learning method by which a model parameter of a learning model used in control of a moving body having movement constraints can be appropriately learned. In this model parameter learning method, a model prediction control algorithm reflecting movement constraints of a robot 1 is used to calculate a time series of learning speed commands such that the movement trajectory of the robot 1 tracks the time series of a movement trajectory of a first pedestrian 5; and a model parameter of a CNN model is learned by an error back propagation method, the CNN model using learning data including the learning speed commands time series as input and outputting a time series of speed commands for a first moving body.

The present disclosure is directed to methods, computer program products, and computer systems for instructing a robot to prepare a food dish by replacing the human chef's movements and actions. Monitoring a human chef is carried out in an instrumented application-specific setting, a standardized robotic kitchen in this instance, and involves using sensors and computers to watch, monitor, record and interpret the motions and actions of the human chef, in order to develop a robot-executable set of commands robust to variations and changes in the environment, capable of allowing a robotic or automated system in a robotic kitchen to prepare the same dish to the standards and quality as the dish prepared by the human chef.

Disclosed herein is a method, system, and non-volatile storage medium for simplifying the automation of a process of flow. The method may include determining a machine-independent process model based on data representing a handling of a work tool for performing a process flow. The process flow may include a plurality of sub-processes and the process model may link a process activity with spatial information for each sub-process. The method may also include mapping the machine-independent process model to a machine-specific control model of a machine using a model of the machine. The machine-specific control model may define an operating point of the machine for each sub-process, and the operating point may correspond to the process activity and to the spatial information.

The invention relates to a method for preparing and carrying out tasks by means of a robot, in which method, in a preparation phase at least one probable action goal is determined on the basis of surroundings information and taking into account a user command probability, at least one preparation action sequence is generated which is aimed at the at least one probable action goal, and the at least one preparation action sequence is carried out. The invention also relates to a robot for carrying out tasks, and to a computer program.

A flexible robotic kitchen assistant automates the preparation of food items. The robotic kitchen assistant includes a robotic arm, a sensor assembly comprising a plurality of cameras aimed at a kitchen workspace for preparing the food items, a controller operable to move the robotic arm, and a processor. The processor is operable to command the robotic arm to perform a food preparation step on the food items in the kitchen workspace based on order information, recipe information, kitchen equipment information, and camera data. The system is capable of performing a wide range of tasks commonly used in residential and commercial kitchens and is able to work collaboratively with and in close proximity to human kitchen workers.

Disclosed herein are embodiments of systems and methods for diversified imitation learning for automated machines. In an embodiment, a process-profiling system obtains sensor data captured by a plurality of sensors that are arranged to observe one or more human subjects performing one or more processes to accomplish one or more tasks. The process-profiling system clusters the sensor data based on a set of one or more process-performance criteria. The process-profiling system also performs, based on the clustered sensor data, one or both of generating and updating one or more process profiles in a plurality of process profiles. The process-profiling system selects, for one or more corresponding automated machines, one or more process profiles from among the plurality of process profiles, and the process-profiling system configures the one or more corresponding automated machines to operate according to the selected one or more process profiles.

A method for controlling a robotic manipulator according to a task comprises accepting a feedback signal including a sequence of multi-modal observations of a state of execution of the task. The multi-modal observations are processed with a neural network having a self-attention module with a hierarchically conditioned output to produce a skill of the robotic manipulator and an action conditioned on the skill. The neural network is trained in a supervised manner with demonstration data to produce a sequence of skills and a corresponding sequence of actions for the actuators of the robotic manipulator to perform the task. The method further comprises determining one or more control commands for the one or more actuators based on the produced action and submitting the one or more control commands to the one or more actuators causing a change of the state of execution of the task.

A method includes determining, for a robotic device that comprises a perception system, a robot planner state representing at least one future path for the robotic device in an environment. The method also includes determining a perception system trajectory by inputting at least the robot planner state into a machine learning model trained based on training data comprising at least a plurality of robot planner states corresponding to a plurality of operator-directed perception system trajectories. The method further includes controlling, by the robotic device, the perception system to move through the determined perception system trajectory.

A robot remote operation control device includes, in robot remote operation control for an operator to remotely operate a robot capable of gripping an object, an information acquisition unit that acquires operator state information on a state of the operator who operates the robot, an intention estimation unit that estimates a motion intention of the operator who causes the robot to perform a motion, on the basis of the operator state information, and a gripping method determination unit that determines a gripping method for the object on the basis of the estimated motion intention of the operator.

A control engine is trained to operate a robotic camera according to a variety of different cinematographic techniques. The control engine may reconfigure the robotic camera to respond to a set of cues, to enforce a set of constraints, or to apply one or more characteristic styles. A training engine trains a network within the control engine based on training data that exemplifies cue responses, enforced constraints, and characteristic styles.