A handling device according to an embodiment includes a manipulator, a normal grid generation unit, a hand kernel generation unit, a calculation unit, and a control unit. The normal grid generation unit converts a depth image into a point cloud, generates spatial data including an object to be grasped that is divided into a plurality of grids from the point cloud, and calculates a normal vector of the point cloud included in the grid using spherical coordinates. The hand kernel generation unit generates a hand kernel of each suction pad. The calculation unit calculates ease of grasping the object to be grasped by a plurality of suction pads based on a 3D convolution calculation using a grid including the spatial data and the hand kernel. The control unit controls a grasping operation of the manipulator based on the ease of grasping the object to be grasped by the plurality of suction pads.

The invention provides a robot system that enables easy, efficient, and precise checking through simulation. The invention includes a virtual model display unit configured to place virtual models in a virtual space on a screen and display the virtual models simultaneously with real equipment; a robot program teaching unit configured to perform teaching of a robot program in the virtual space; a real space virtual model display unit configured to display the virtual models and teaching points of the robot program in a real space, based on a positional relationship in the virtual space; and a virtual model placement position correcting unit configured to correct placement positions of the virtual models to match the real equipment in the real space.

Disclosed herein are an apparatus and method for generating robot interaction behavior. The method for generating robot interaction behavior includes generating co-speech gesture of a robot corresponding to utterance input of a user, generating a nonverbal behavior of the robot, that is a sequence of next joint positions of the robot, which are estimated from joint positions of the user and current joint positions of the robot based on a pre-trained neural network model for robot pose estimation, and generating a final behavior using at least one of the co-speech gesture and the nonverbal behavior.

Provided is a robot that includes: a first sensor having a first output and configured to sense state of a robot or an environment of the robot; a first hardware machine-learning accelerator coupled to the first output of the first sensor and configured to transform information sensed by the first sensor into a first latent-space representation; a second sensor having a second output and configured to sense state of the robot or the environment of the robot; a second hardware machine-learning accelerator configured to transform information sensed by the second sensor into a second latent-space representation; and a processor configured to control the robot based on both the first latent-space representation and the second latent-space representation.

A method of the present disclosure includes (a) receiving settings of an objective function and a constraint condition, (b) controlling a robot to execute work using a candidate value of a control parameter and measuring a performance index value for the objective function and a constraint evaluation value, (c) searching for a next candidate value of the control parameter by executing optimization processing using a value of the objective function, (d) obtaining the values of the objective function and the constraint evaluation values with respect to the plurality of candidate values by repeating (b) and (c), and (e) displaying a processing result containing a correlation chart showing the values of the objective function and the constraint evaluation values with respect to each of the plurality of candidate values.

A force control parameter setup support method of supporting a setup of a force control parameter to be used for force control when controlling a robot arm a tip of which is attached with a polishing tool using the force control to perform a polishing task on an object including a first step of obtaining task information related to the polishing task, a second step of selectively reading out information of the force control parameter corresponding to the task information obtained in the first step from a storage section in which a plurality of pieces of information of the force control parameter is stored, and a third step of displaying the information of the force control parameter read out in the second step on a display section.

A modular frame for an intelligent robot includes a base and one or more devices for performing specified functions. The base controls the actions and functions of the robot and contains a memory of operating instructions for a plurality of modules, each module performing unique functions. The base has a smart connector. The devices for performing specified functions have a smart connector which contains a unique code for that device or module and firmware for operation of the module. When a module is affixed on the modular frame, the smart connector of the module electronically communicates with the smart connector of the base, thereby providing the base with sufficient operating information to operate the intelligent robot.

In current applications of autonomous machines in industrial settings, the environment, in particular the devices and systems with which the machine interacts, is known such that the autonomous machine can operate in the particular environment successfully. Thus, current approaches to automating tasks within varying environments, for instance complex environments having uncertainties, lack capabilities and efficiencies. In an example aspect, a method for operating an autonomous machine within a physical environment includes detecting an object within the physical environment. The autonomous machine can determine and perform a principle of operation associated with a detected subcomponent of the object, so as to complete a task that requires that the autonomous machine interacts with the object. In some cases, the autonomous machine has not previously encountered the object.

A method for controlling a robotic system includes: capturing, by an imaging system, one or more images of a scene; computing, by a processing circuit including a processor and memory, one or more instance segmentation masks based on the one or more images, the one or more instance segmentation masks detecting one or more objects in the scene; computing, by the processing circuit, one or more pickability scores for the one or more objects; selecting, by the processing circuit, an object among the one or more objects based on the one or more pickability scores; computing, by the processing circuit, an object picking plan for the selected object; and outputting, by the processing circuit, the object picking plan to a controller configured to control an end effector of a robotic arm to pick the selected object.

Utilizing an initial set of offline positive-only robotic demonstration data for pre-training an actor network and a critic network for robotic control, followed by further training of the networks based on online robotic episodes that utilize the network(s). Implementations enable the actor network to be effectively pre-trained, while mitigating occurrences of and/or the extent of forgetting when further trained based on episode data. Implementations additionally or alternatively enable the actor network to be trained to a given degree of effectiveness in fewer training steps. In various implementations, one or more adaptation techniques are utilized in performing the robotic episodes and/or in performing the robotic training. The adaptation techniques can each, individually, result in one or more corresponding advantages and, when used in any combination, the corresponding advantages can accumulate. The adaptation techniques include Positive Sample Filtering, Adaptive Exploration, Using Max Q Values, and Using the Actor in CEM.