Abstract: An object gripping method includes determining a gripping center coordinate system when gripping a target object assumed to be gripped by an end effector having a plurality of fingers for each of a plurality of gripping postures that are able to be taken by the end effector, determining an initial fingertip position from the determined gripping center coordinate system, instructing the end effector to grip the target object at the initial fingertip position and gripping the target object, fixing a fingertip position of the end effector with respect to the gripping center coordinate system, and determining an operation amount of the gripping center coordinate system according to a desired operation amount of the target object and operating the target object according to an operation of the gripping center coordinate system.

Abstract: An object holding method includes: a step of determining a center of a first object assumed to be held by an end effector for a plurality of holding postures which the end effector is able to take; a step of measuring a shape of a second object based on an image in which the second object is captured from a viewpoint; a step of selecting a holding posture when the end effector holds the second object from the holding postures based on centers of the second object and the first object; and a step of reducing a portion of the shape of the second object in a case where the end effector holds the second object by using the selected holding posture.

Abstract: The disclosure provides a gripping position determination device, a gripping position determination system, a gripping position determination method, and a recording medium. The gripping position determination device for a robot hand having a plurality of multi joint fingers includes: a frictional force distribution calculation part estimating, from a predictive control of a gripping force when an object is gripped by at least two fingers, a frictional force between one of the gripping fingers and the object, and calculates a frictional force distribution where grapping of the object is possible on a surface of the object based on a value related to a frictional force calculated by using the estimated frictional force; a grippable region selection part selecting, from the frictional force distribution, at least one grippable region; and a gripping position calculation part calculating, from the selected grippable region, a gripping position where stable gripping of the object is possible.

Abstract: It is possible to perform robot motor learning in a quick and stable manner using a reinforcement learning apparatus including: a first-type environment parameter obtaining unit that obtains a value of one or more first-type environment parameters; a control parameter value calculation unit that calculates a value of one or more control parameters maximizing a reward by using the value of the one or more first-type environment parameters; a control parameter value output unit that outputs the value of the one or more control parameters to the control object; a second-type environment parameter obtaining unit that obtains a value of one or more second-type environment parameters; a virtual external force calculation unit that calculates the virtual external force by using the value of the one or more second-type environment parameters; and a virtual external force output unit that outputs the virtual external force to the control object.

Abstract: A control method for a legged mobile robot includes exercising a body of a robot such that a center of gravity of the robot obtains a momentum or the body obtains an angular momentum in a direction in which an object is to be moved while restraining a force from being applied to the object from the robot in a state wherein the robot opposes the object, and applying a force to the object from a hand of an arm body provided in the body of the robot so as to start moving the object in a state wherein the center of gravity has obtained the momentum or the body has an angular momentum. With this arrangement, when moving an object by a robot, a motion of the robot can be smoothly changed while preventing a significant change in ZMP before and after starting to move the object.

Abstract: A robot and a behavior control system for the same are capable of ensuring continued stability while carrying out a specified task by a motion of a body of the robot. Time-series changing patterns of first state variables indicating a motional state of an arm are generated according to a stochastic transition model such that at least one of the first state variables follows a first specified motion trajectory for causing the robot to carry out a specified task. Similarly, time-series changing patterns of second state variables indicating a motional state of the body are generated according to the stochastic transition model such that the second state variables satisfy a continuously stable dynamic condition.

Abstract: A reinforcement learning system (1) of the present invention utilizes a value of a first value gradient function (dV1/dt) in the learning performed by a second learning device (122), namely in evaluating a second reward (r2(t)). The first value gradient function (dV1/dt) is a temporal differential of a first value function (V1) which is defined according to a first reward (r1(t)) obtained from an environment and is served as a learning result given by a first learning device (121). An action policy which should be taken by a robot (R) to execute a task is determined based on the second reward (r2(t)).

Abstract: A controller of a mobile robot that moves an object such that the position of a representative point of the object and the posture of the object follow a desired position and posture trajectory is provided. The desired posture trajectory of the object includes the desired value of the angular difference about a yaw axis between a reference direction, which is a direction orthogonal to the yaw axis of the object, and the direction of the moving velocity vector of the representative point of the object, defined by the desired position trajectory. The controller has a desired angular difference setting means which variably sets the desired value of the angular difference according to at least a required condition related to a movement mode of the object. This allows the object to be moved at a posture which meets the required condition of the movement mode.

Abstract: The present invention provides a motion control system to control a motion of a second motion body, by considering an environment which a human contacts and a motion mode appropriate to the environment, and an environment which a robot actually contacts. The motion mode is learned based on an idea that it is sufficient to learn only a feature part of the motion mode of the human without a necessity to learn the others. Moreover, based on an idea that it is sufficient to reproduce only the feature part of the motion mode of the human without a necessity to reproduce the others, the motion mode of the robot is controlled by using the model obtained from the learning result. Thereby, the motion mode of the robot is controlled by using the motion mode of the human as a prototype without restricting the motion mode thereof more than necessary.

Abstract: It is possible to perform robot motor learning in a quick and stable manner using a reinforcement learning apparatus including: a first-type environment parameter obtaining unit that obtains a value of one or more first-type environment parameters; a control parameter value calculation unit that calculates a value of one or more control parameters maximizing a reward by using the value of the one or more first-type environment parameters; a control parameter value output unit that outputs the value of the one or more control parameters to the control object; a second-type environment parameter obtaining unit that obtains a value of one or more second-type environment parameters; a virtual external force calculation unit that calculates the virtual external force by using the value of the one or more second-type environment parameters; and a virtual external force output unit that outputs the virtual external force to the control object.

Abstract: A state estimating apparatus permits efficient, highly accurate estimation of the state of an object. A particle in a state variable space defined by a second state variable preferentially remains or increases as the likelihood thereof relative to a current measured value of a first state variable is higher, while a particle is preferentially extinguished as the likelihood thereof is lower. A particle which transitions in the state variable space according to a state transition model with a high probability of being followed by an object (a high-likelihood model) as a next model tends to increase. On the other hand, although in a small quantity, there are particles having models (low-likelihood models) which are different from the high-likelihood model as their unique models.

Abstract: A provisional desired motion trajectory of an object is determined based on a moving plan of the object. Then, it is determined whether a robot leg motion can satisfy a necessary requirement. The requirement is related to a position/posture relationship between the object and the robot, and a determination of whether the requirement can be satisfied is made at a future, predetermined step. A restrictive condition related to robot leg motion is satisfied at each step up to the predetermined number of steps. If the requirement is satisfied, then a desired gait is generated on the basis of the provisional desired motion trajectory. Otherwise, a desired gait is generated on the basis of a desired motion trajectory of the object according to a corrected moving plan.

Abstract: A robot and a behavior control system for the same are capable of ensuring continued stability while carrying out a specified task by a motion of a body of the robot. Time-series changing patterns of first state variables indicating a motional state of an arm are generated according to a stochastic transition model such that at least one of the first state variables follows a first specified motion trajectory for causing the robot to carry out a specified task. Similarly, time-series changing patterns of second state variables indicating a motional state of the body are generated according to the stochastic transition model such that the second state variables satisfy a continuously stable dynamic condition.

Abstract: In a mobile robot control system, it is configured such that the robot generates time-series data sequentially at a predetermined time interval and transmits them to the external terminal, and the external terminal receives the transmitted time-series data and adds them to the motion command, such that the motion of the robot is determined based on the generated time-series data and the time-series data added to the motion command. With this, it becomes possible to prevent the robot from suddenly starting to move at the time when the communication between the external terminal which is a transmitting source of the motion command and the robot has recovered from disconnection, thereby enabling to avoid making the operator feel unnatural.

Abstract: A gait generating system of a legged mobile robot is provided with a device for determining a desired trajectory of an external force to be applied to a robot 1, a device for determining a parameter of a desired gait (current time gait) for a predetermined period on the basis of a desired trajectory of an external force or the like, a device for determining a parameter of a virtual cyclic gait that follows the current time gait on the basis of the desired trajectory of the external force or the like, a device for correcting the current time gait parameter such that a body motion trajectory of the robot 1 of the current time gait converges to a body motion trajectory of the cyclic gait, and a device for sequentially determining an instantaneous value of the current time gait on the basis of the corrected current time gait parameter.

Abstract: In a mobile robot control system, it is configured such that the robot generates time-series data sequentially at a predetermined time interval and transmits them to the external terminal, and the external terminal receives the transmitted time-series data and adds them to the motion command, such that the motion of the robot is determined based on the generated time-series data and the time-series data added to the motion command. With this, it becomes possible to prevent the robot from suddenly starting to move at the time when the communication between the external terminal which is a transmitting source of the motion command and the robot has recovered from disconnection, thereby enabling to avoid making the operator feel unnatural.

Abstract: A controller of a leg type moving robot determines an action force to be input to an object dynamic model 2 such that a motion state amount (object model velocity) of the object dynamic model 2 follows a desired motion state amount based on a moving plan of an object, and also determines a manipulated variable of the motion state amount (object model velocity) of the object dynamic model 2 such that the difference between an actual object position and a desired object position approximates zero, and then inputs the determined action force and manipulated variable to the object dynamic model 2 to sequentially determine the desired object position. Further, a desired object reaction force to a robot from the object is determined from the determined reaction force.

Abstract: When a new desired gait of a robot is generated, it is determined, on the assumption that the trajectory of an acting force between the robot and an object at a predetermined time point in the future changes to a trajectory different from a desired trajectory, whether a predetermined dynamical restrictive condition can be satisfied when a desired gait after the predetermined time point is generated. If the condition cannot be satisfied, then a moving schedule for the object is corrected, the desired trajectory or the like of the acting force between the robot and the object is re-determined, and a new desired gait is generated using the re-determined desired trajectory. With this arrangement, the gait of the robot to cause the robot to perform an operation for moving the object is generated such that the stability of the posture of the robot can be secured even if an acting force between the robot and the object in the future deviates from a desired value.

Abstract: A gait generator determines a desired motional trajectory and a desired object reaction force trajectory of an object 120 for a predetermined period after the current time by using an object dynamic model while supplying, to the object dynamic model, a model manipulated variable (estimated disturbance force) for bringing a behavior of the object 120 on the object dynamic model close to an actual behavior, and provisionally generates a gait of a robot 1 for a predetermined period by using the aforesaid determined trajectories. Based on the gait and an object desired motion trajectory, a geometric restrictive condition, such as interference between the robot 1 and the object 120, is checked, and a moving plan for the object 120 or a gait parameter (predicted landing position/posture or the like) of the robot 1 is corrected as appropriate according to a result of the check, so as to generate a gait of the robot 1.

Abstract: A reinforcement learning system (1) of the present invention utilizes a value of a first value gradient function (dV1/dt) in the learning performed by a second learning device (122), namely in evaluating a second reward (r2(t)). The first value gradient function (dV1/dt) is a temporal differential of a first value function (V1) which is defined according to a first reward (r1(t)) obtained from an environment and is served as a learning result given by a first learning device (121). An action policy which should be taken by a robot (R) to execute a task is determined based on the second reward (r2(t)).