ROBOT CONTROL DEVICE, ROBOT CONTROL METHOD AND COMPUTER MEMORY READABLE MEDIUM

Abstract

Robot control device (10), comprising: a process time prediction unit (11) for predicting a robot process time (T1), which is a time required by a robot (30) to cause a hand (31) of the robot (30) to reach a target position (P) get ) achieved, and a manual operation time (T gsp ), where this is a time from the time at which an action of the hand (31) is commanded until the time at which the hand (31) carries out an action of grasping an object at the target position (P get ) finished, acts; and a hand operation start instruction unit (13) for issuing an instruction to start an operation of the hand (31) at a time determined based on the predicted robot operation time (T1) and the predicted hand operation time (T gsp ) is determined, whereby the process time prediction unit (11) predicts the robot process time (T1) by calculation, which integrates the robot control system information, which includes a control parameter to determine the responsiveness of a control system of the robot (30).

Description

[0001] The present disclosure relates to a robot control device that controls a robot, a robot control method and a robot control program.

[0002] A robot is known that grasps a substance using a gripper attached to a distal end section of the robot, i.e., a hand, and transports the grasped substance to a specific position. Such a robot is used for assembly work or for removing a workpiece from a machine tool in the electrical and electronics industry, for washing operations in the food industry, for transfer work in the logistics industry, and similar applications. Hereinafter, a substance that can be grasped by the hand is referred to as an object. In robotic conveying operations, the robot is first positioned at the location of an object, which is detected using a measuring device, or at a position predetermined by a program. When the hand then executes a command, it initiates a process.After a predetermined waiting period has elapsed since the hand grasped the object, the robot starts the next process. With such a series of grasping operations, it is desirable to optimize the timing of the hand's command or the length of the waiting period in order to reduce the overall process time required for both the robot and the hand to grasp the object.

[0003] Patent document 1 discloses a robot control device that initiates a hand operation before the robot reaches a target position, which is the position of an object. The robot control device from patent document 1 initiates a hand operation when the time it takes for the robot to reach the target position becomes shorter than the hand operation time, which is the time required for the hand's operation. Specifically, the robot control device from patent document 1 predicts a movement time from the robot's current position to its target position based on a robot operation command for controlling the robot and compares the predicted movement time with the hand operation time. The robot control device from patent document 1 issues a command to the hand when the predicted movement time becomes shorter than the hand operation time.

[0004] Patent literature 2 discloses an operating setting device with a camera arranged to capture a robot and a hand. Additionally, a robot control device comprises an operating control unit that sends operating commands to the robot according to a predefined control cycle. The camera captures images at time intervals corresponding to the control cycle. A determination unit of the robot control device determines, based on the results of image processing by an image processing unit, whether the hand's operation is appropriate or not. If the determination unit determines that the hand's operation is not appropriate, a correction unit of the robot control device corrects the command text contained in an operating program so that the hand's operation becomes the appropriate operation.

[0005] Patent literature 3 discloses a machine learning device for learning the operating state of a robot that stores a plurality of objects arranged on a carrier device in a container using a hand to grasp the objects. A machine learning method is described that learns an optimal object grasping path when a robot grasps objects arranged on a carrier device. Patent literature 4 discloses an article transfer device and an article transfer method for transferring articles using a robot configured to control the robot using information about the article detected by the image processing area.

[0006] Patent literature 5 discloses a command generation device which provides a command to a control device for starting and controlling an actuator in order to start a machine, in particular a machine with low stiffness.

[0007] Patent literature 6 discloses a robot control method in which each servo motor is controlled such that the tip of the robot is moved from a first position to a second position, based on the detection angles of a reduction gear and the inclination of an output shaft, as detected by each output encoder. A time duration is determined that elapses from the time at which the detection angles of each input encoder reach the target angles until the time at which a pivot amplitude at the position of the robot tip in the calculated second position converges within a predetermined range.

[0008] Patent literature 7 discloses a workpiece removal device that calculates a gripping operation to reduce errors when removing a workpiece from a multitude of workpieces and also to reduce repetition operations using sensors and information about the success or failure of previous removals. Patent literature 1: Japanese publication JP 2000 - 000 787 A Patent literature 2: German patent application DE 10 2020 107 385 A1 Patent literature 3: German patent application DE 10 2017 008 836 B4 Patent literature 4: German patent application DE 10 2015 015 638 B4 Patent literature 5: German patent application DE 10 2005 059 530 B4 Patent Literature 6: Japanese Publication JP 2020 - 062 730 A Patent literature 7: Japanese publication JP 2013 - 052 490 A

[0009] The technical problem arising from the current state of the art is that the time it takes for the robot to move its hand to the target position is not solely determined by the robot's action command, but also varies due to a delay in the robot's operation caused by characteristics of the robot's control system. The length of this delay varies depending on the speed or acceleration of each of the robot's axes. In a case where a parameter of the robot's control system is made variable depending on the robot's position or orientation, the length of the delay will also vary with each change in the robot's position or orientation. The movement time predicted based on the robot's action command will therefore have an error compared to the actual movement time.Therefore, determining the precise moment to issue an instruction for the hand's action requires a skilled person to perform adjustments over a long period. As described above, according to the conventional technique disclosed in patent literature 1, a problem arises in that it is difficult to precisely adjust the timing for issuing an instruction for the hand's action in order to shorten the time required for the grasping process to seize the object.

[0010] The present disclosure was made in light of the foregoing, and one of its objectives is to provide a robot control device which, in order to reduce the time required for a grasping operation by a robot and a hand, can accurately set a time for issuing an instruction for an operation of the hand.

[0011] To solve the above problem and fulfill a task, a robot control device according to claims 1, 2 and 3 comprises a computer-readable storage medium according to claim 15, and a robot control method according to claim 14.

[0012] A robot control device according to the present disclosure achieves an effect which, in order to shorten the time required for a gripping operation by a robot and a hand, makes it possible to precisely set a time for giving an instruction for an operation of the hand. Fig. Figure 1 is a representation illustrating an exemplary configuration of a robot control device according to a first embodiment. Fig. Figure 2 is a representation illustrating an exemplary configuration of a robot control system that includes the robot control device according to the first embodiment. Fig. Figure 3 is a representation to explain a robot operation time predicted by an operation time prediction unit of the robot control device, according to the first embodiment. Fig. Figure 4 is a representation to explain an instruction to start an operation of a hand and an instruction to start an operation of a robot by the robot control device according to the first embodiment. Fig. Figure 5 is a flowchart illustrating a process procedure of the robot control device according to the first embodiment. Fig. Figure 6 is a representation illustrating an exemplary configuration of a robot control device according to a second embodiment. Fig. Figure 7 is a representation illustrating an exemplary configuration of a robot control device according to a third embodiment. Fig. Figure 8 is a representation illustrating an exemplary configuration of a robot control device according to a fifth embodiment. Fig. Figure 9 is a representation illustrating an exemplary configuration of a robot control device according to a sixth embodiment. Fig. Figure 10 is a representation illustrating a learning device and a storage unit for learned models in a gripper control parameter learning unit included in the robot control device according to the sixth embodiment. Fig. Figure 11 is a representation illustrating an exemplary configuration of a neural network used for machine learning in the sixth embodiment. Fig. Figure 12 is a representation illustrating an inference device and the storage unit for learned models in the gripper control parameter learning unit, which is included in the robot control device according to the sixth embodiment. Fig. Figure 13 is a representation illustrating an exemplary configuration of a robot control device according to a seventh embodiment. Fig. Figure 14 is a representation illustrating an exemplary configuration of a control circuit according to the first to seventh embodiments. Fig. Figure 15 is a representation illustrating an exemplary configuration of a hardware circuit as a dedicated circuit according to the first to seventh embodiments.

[0013] The following section describes in detail, with reference to the drawings, a robot control device, a robot control method and a robot control program according to each embodiment.

[0014] A first embodiment is described in more detail below.

[0015] Fig. Figure 1 is a representation illustrating an exemplary configuration of a robot control device 10 according to a first embodiment. Fig. Figure 2 is a representation illustrating an exemplary configuration of a robot control system 40, which includes the robot control device 10 according to the first embodiment. The robot control system 40 includes the robot control device 10, a robot 30, and a hand 31. The hand 31 is attached to a distal end section of an arm representing the robot 30. The robot control device 10 controls the robot 30. The robot control system 40 grasps an object and moves the grasped object into a defined position through actions of the robot 30 and the hand 31. Fig. Figure 2 illustrates robot control system 40, which removes an object placed in a box and transports it to a predetermined position outside the box. A sensing device, such as an image sensor, is installed as a peripheral device of robot control system 40. The sensing device detects the object's position. Robot control 10 receives information about the object's position from the sensing device and generates a robot operation command based on this information. Robot control 10 sends this robot operation command to a drive unit of robot 30. The drive unit is not shown. The drive unit propels robot 30 according to the robot operation command, causing robot 30 to operate as directed.

[0016] Hand 31 is controlled by a hand control device. The hand control device is not illustrated. Hand 31 operates according to a hand operation command sent by the hand control device. In the robot control system 40, the robot control device 10 can control both the robot 30 and hand 31 without using the hand control device.

[0017] In the first embodiment, the hand 31 comprises two movable units that perform opening and closing operations. The hand 31 grasps an object by grasping the object with two movable units facing each other. The hand 31 grasps the object by a closing operation in which the movable units move in a direction that brings them closer together. The hand 31 releases the grasped object by an opening operation in which the movable units move away from each other from a state of grasping the object. It should be noted that the hand 31 is not limited to one that comprises two movable units. The number of movable units included in the hand 31 is not limited to two and can be any number. The hand 31 is not limited to one that comprises such movable units.For example, hand 31 can be one that includes a section which generates suction and grasps an object by causing the section to adsorb the object. In the following description, the movable units of hand 31 are each referred to as a finger.

[0018] As in Fig. As illustrated in Figure 1, the robot control device 10 includes an operation time prediction unit 11, an adjustment time prediction unit 12, a manual operation start instruction unit 13, a re-reaching operation instruction unit 14, and a memory unit 15. Additionally, the robot control device 10 includes an instruction generation unit that generates a robot operation instruction. The instruction generation unit is not shown.

[0019] The process time prediction unit 11 includes a robot process time calculation unit 16, which calculates a robot process time T1, and a manual process time calculation unit 17, which calculates a manual process time T gsp calculated. The robot operation time T1 is the time that robot 30 needs to cause the hand 31 of robot 30 to reach a target position. The hand operation time T gspis the time from the moment the operation of hand 31 is commanded until the moment hand 31 completes an operation of grasping the object at the target position. By calculating the robot operation time T1, the robot operation time prediction unit 11 predicts the robot operation time T1. The robot operation time calculation unit 16 calculates a remaining robot operation time T0 based on the robot operation time T1. The remaining robot operation time T0 is the time required from any point after the start of the operation of robot 30 until the time when robot 30 causes hand 31 to reach the target position. By calculating the hand operation time T1, the hand operation time calculation unit 17 predicts the robot operation time T1. gsp The process time prediction unit 11 calculates the manual process time T. gsp previously.

[0020] The process time prediction unit 11 provides a value for the remaining robot process time T0 and a value for the manual process time T. gsp to the hand operation start instruction unit 13. The hand operation start instruction unit 13 determines a time to start the operation of hand 31 using the value of the remaining robot operation time T0, which is based on the value of the robot operation time T1 and the value of the hand operation time T. gsp The hand operation start instruction unit 13 instructs the hand control device to start the operation of hand 31 at the specified time. That is, the hand operation start instruction unit 13 issues an instruction to start the operation of hand 31 at the time determined based on the predicted robot operation time T1 and the predicted hand operation time T1. gspis determined. In addition, the hand operation start instruction unit 13 outputs information to the follow-up operation instruction unit 14, specifying a time Tk at which the instruction to start the operation of hand 31 is given.

[0021] The adjustment time prediction unit 12 predicts an adjustment time T. fit before. The adjustment time T fit Adaptation time is the time from when hand 31 starts grasping the object until hand 31 "fits" the object. "Hand 31 fits" means that the contact between hand 31 and the object becomes stable enough that hand 31 can maintain its grip even when robot 30 moves hand 31. The adaptation time prediction unit 12 provides a value for the predicted adaptation time T. fitto the re-engaging process instruction unit 14. In the first embodiment, the length of the predicted adjustment time T is fit The length of the waiting time is defined as the time from when hand 31 grasps the object until robot 30 starts the next operation. After the predicted adaptation time T has elapsed, fitOnce hand 31 begins grasping the object, the re-grasping instruction unit 14 issues an instruction for the robot 30 to perform the action following the object grasping process. Information about the object's position P and hand action information is input into the robot control unit 10. The position P is measured using a sensor, such as an image sensor. The position P information to be input into the robot control unit 10 is updated each time the object is grasped. The hand action information is information about the action of hand 31 in grasping the object. In the first embodiment, the hand action information includes values ​​for the insertion depth d, the opening width w, and the gripping force F. gsp .

[0022] Before the closing process of grasping the object begins, a distal end of hand 31, that is, a distal end of each finger, is inserted around the object. The extent of insertion d represents the degree of insertion of each finger relative to the object. Here, a state is defined as d=0, where, with the object's midpoint as the reference, the distal end of each finger coincides with the object's midpoint. "The distal end of each finger coincides with the object's midpoint" means that the position of each finger's distal end in an insertion direction and the object's midpoint in that insertion direction coincide. The insertion direction is the direction in which each finger is moved by the robot 30's action prior to the closing process.Additionally, a state is defined as d>0, where the object's central position is located on a base side of each finger of hand 31, compared to the case of d=0. A state is defined as d<0, where the object's central position is located on a distal end side of each finger, compared to the case of d=0. The opening width w is an interval between the distal end sections of the respective fingers at the start of the closing process. The gripping force F. gsp is a force that hand 31 exerts on the object in a state in which hand 31 grasps the object.

[0023] Memory unit 15 stores robot control system information, hand property information, and object property information. The robot control system information specifies the properties of a robot control system that controls the robot's drive unit 30. This information includes at least one value of a control parameter used for forward-feedback control by the robot control system and one value of a control parameter used for backward-feedback control. The control parameter included in the robot control system information determines the responsiveness of the robot control system.A control parameter is, for example, a parameter defined in control engineering, such as a proportional (P) gain, an integral (I) gain, and a differential (D) gain of a proportional-integral-differential (PID) controller, a time constant of a first-order delay element, or a dead time of a dead-time element. The type of control parameter included in the robot control system information varies depending on the configuration of the robot control system.

[0024] The hand property information is information that specifies the operational properties of hand 31. This information includes at least one value for an opening / closing command speed, hand control system information, and a value for a mechanism parameter. The opening / closing command speed is a speed commanded by the hand control device and represents the movement speed of each finger during the opening and closing process of hand 31. The hand control system information specifies the properties of a hand control system that controls the opening and closing of hand 31. This information includes a value for a control parameter of the hand control system.

[0025] The mechanism parameter is a value that expresses a positional relationship of an element, such as an actuator driving hand 31 or a reduction gear, or a dimension of each segment of hand 31's finger. Alternatively, the mechanism parameter is a value that expresses a reduction ratio of an actuator or a reduction gear provided at hand 31. The actuator provided at hand 31 is a motor or a pneumatic actuator. Using the mechanism parameter makes it possible to calculate the position of hand 31 or the position of each finger. The mechanism parameter is used to calculate the time at which hand 31 comes into contact with the object or the time at which hand 31 comes into contact with a substance surrounding the object.

[0026] To improve the stability of the grip by hand 31, a flexible material can be provided in a finger rest section, which is a segment of each finger of hand 31 that contacts the object. In this case, the object requires time to conform to the finger rest section through deformation, from the moment the object begins to make contact with the finger rest section until the object is gripped stably. To account for such properties of the finger rest section, the hand property information can include information about the physical properties of the finger rest section, which is a segment of each finger of hand 31 that contacts the object.The information about the physical properties of the finger support section includes a value of the stiffness Kh of the finger support section, a value of the viscosity Dh of the finger support section, or the like.

[0027] Object property information is information that specifies the physical properties of the object. This includes information about the object's deformation and contains values ​​such as stiffness (Kw), viscosity (Dw), or similar parameters.

[0028] One in Fig. 2 illustrated position P get is a target position at which the distal end of hand 31 is reached by the robot 30. Information about the position P of the object to be grasped next is entered into the robot control device 10. The position P getis determined based on the object's position P and the extent of the insertion d. A in Fig. 2 illustrated position P up is a position of hand 31 when the process of the robot 30 to move the distal end of hand 31 into the target position is started.

[0029] The detection device, which is a peripheral device of the robot control system 40, calculates the object's position P and the extent of insertion d based on the result of observing the object. The extent of insertion d calculated here is the extent of insertion d in a state where the distal end of the hand 31 is caused to reach position P get The command generation unit of the robot control device 10 generates a robot operation command based on the information about the object's position P and the insertion depth d. The position P upis a position that is a certain distance from position P get is spaced upwards, and is a position that can be defined by a user. The distance between position P up and position P get A distance is established that prevents the distal end of hand 31 from touching things around the object. This is described as a position of hand 31 at position P. up The same position is established as the position that hand 31 has at position P. get occupies position P up and the position at position P up As defined above, a relative positional relationship between the object and the hand 31 can be made the same in each of the gripping operations, regardless of the operation of the robot 30, immediately before the hand 31 reaches position P upThis is achieved. Since the relative positional relationship between the object and hand 31 is always the same, the robot 30 can execute a reproducible process. The robot process command generated here is a robot process command to cause the robot 30 to perform a process of moving hand 31 from position P. up to position P get carries out, with position P get based on the position P of the object and the extent of the insertion d, and the position P up as defined above. As described above, the robot operation command is generated based on the information about the object's position P and the value of the insertion extent d, which is the hand operation information. Next, an operation of the robot control device 10 is described. Fig. Figure 3 is a diagram illustrating the robot operation time T1, which is predicted by the operation time prediction unit 11 of the robot control device 10, according to the first embodiment. The robot operation time T1 is an operation time of the robot 30 in a case where the robot 30 performs the operation of moving the hand 31 from position P. up to position P get The robot performs the operation according to the robot action command. A start point of the robot action time T1 is a time at which the robot 30 starts the operation. An end point of the robot action time T1 is a time at which an error occurs between the position of the distal end of the hand 31 and position P. get falls within a predefined range. The robot operation time calculation unit 16 calculates the robot operation time T1 based on the robot operation command.

[0030] A time Ta is the elapsed time since the robot 30 started the operation. The robot operation time calculation unit 16 measures the time Ta. The robot operation time calculation unit 16 calculates the remaining robot operation time T0 by subtracting the time Ta from the robot operation time T1. The operation time prediction unit 11 calculates the remaining robot operation time T0, thereby grabbing, if necessary, a time required for the distal end of the hand 31 to reach the object.

[0031] The robot process time calculation unit 16 includes a robot process simulation tool. This tool simulates the generation of a robot process command and the robot's operation 30 based on the properties of the robot control system. The robot process simulation tool acquires information about the object's position P, which is input into the robot control device 10. It also acquires the insertion depth d value from the manual process information, which is also input into the robot control device 10. Based on the object's position P and the insertion depth d, the robot process simulation tool simulates the generation of the robot process command. Finally, the robot process simulation tool includes a filter for simulating the robot's operation 30 based on the properties of the robot control system.The filter medium is a mathematical model used to simulate the behavior or response of a control system and a mechanical system in response to a command value and to express the state of that behavior or response. A transfer function, for example, can be used as the filter medium. By inputting the command value into the filter medium, the filter medium can obtain, with respect to a part of the mechanical system moving according to the command value, the magnitude of displacement due to an operation as a function of the command value, the magnitude of change in velocity during the operation as a function of the command value, or the magnitude of change in acceleration during the operation as a function of the command value.

[0032] The filter medium simulates the robot 30's operation based on the robot control system information stored in memory unit 15. The robot operation simulation medium simulates the robot 30's operation depending on the robot control system's properties by routing the robot operation command through the filter medium. It should be noted that the robot operation simulation medium and the filter medium are not illustrated.

[0033] The robot operation time calculation unit 16 predicts the actual behavior of the distal end of the hand 31 based on the result of the robot operation simulation by the robot operation simulation means. The robot operation time calculation unit 16 predicts the endpoint of the robot operation time T1 based on the predicted behavior of the hand 31 in order to calculate the robot operation time T1. As described above, the robot operation time calculation unit 16 calculates the robot operation time T1 by means of a calculation that reflects the robot control system information. The operation time prediction unit 11 predicts the robot operation time T1 by means of a calculation that integrates the robot control system information.

[0034] The robot process time calculation unit 16 is not limited to one that simulates the robot's process 30 depending on the properties of the robot control system using the robot process simulation means. The robot process time calculation unit 16 can be one that includes a delay time T. rs The robot's operation time 30 is calculated as a function of the robot control system's properties. The robot operation time calculation unit 16 uses an approximation function that calculates the delay time T. rsThe robot operation time calculation unit 16 calculates the delay time T based on a command position specified by the robot operation command and a velocity or acceleration at which the robot 30 is operated. By analyzing the behavior of the robot control system under a variety of operating conditions simulated in advance, a value for a parameter of the approximation function can be identified based on a simulation result. The value of the parameter of the approximation function is included in the robot control system information stored in memory unit 15. The robot operation time calculation unit 16 calculates the delay time T. rs using the approximation function based on the value of the approximation function parameter contained in the robot control system information. A polynomial equation or a neural network can be used for the approximation function.

[0035] The robot operation time calculation unit 16 calculates a time T1c based on the robot operation command. The time T1c is an operation time of the robot 30 in a case where the operation of moving the hand 31 from position P up to position P get The robot operation time calculation unit 16 calculates the robot operation time T1 by adding the delay time T. rsat time T1c. As described above, the robot operation time calculation unit 16 calculates the robot operation time T1 by a calculation that integrates the robot control system information. That is, the operation time prediction unit 11 predicts the robot operation time T1 by a calculation that integrates the robot control system information. Additionally, the robot operation command used to calculate the robot operation time T1 is generated based on the information about the object's position P and the value of the insertion extent d, which is the manual operation information, as described above. It can be said that the robot operation time calculation unit 16 calculates the robot operation time T1 by a calculation that integrates the robot control system information, the information about the object's position P, and the manual operation information.It can be said that the operation time prediction unit 11 predicts the robot operation time T1 by calculation, which integrates the robot control system information, the information about the object's position P, and the hand operation information. The hand operation time calculation unit 17 includes a hand operation simulation unit. The hand operation simulation unit simulates the hand operation command and simulates the operation of the hand 31 depending on the properties of the hand control system. The hand operation simulation unit extracts a value for an opening width w from the hand operation information that is input into the robot control device 10. The hand operation simulation unit extracts a constant, which is a mechanism parameter, from the hand property information that is stored in the memory unit 15.The hand operation simulation tool calculates the stroke of hand 31 by multiplying the value of the opening width w by the constant. The stroke of hand 31 is the amplitude of movement of the distal end of hand 31 in each of the opening and closing operations. The hand operation simulation tool simulates the generation of the hand operation command based on the calculated stroke.

[0036] The manual operation simulation means includes a filter means for simulating the operation of hand 31 depending on the properties of the hand control system. The filter means simulates the operation of hand 31 based on the hand control system information in the hand property information stored in memory unit 15. The manual operation simulation means simulates the operation of hand 31 depending on the properties of the hand control system by routing the manual operation command through the filter means. It should be noted that the manual operation simulation means and the filter means are not illustrated.

[0037] The manual process time calculation unit 17 calculates a time T gsp_o based on the result of the simulation of the process of each finger of hand 31 by the hand process simulation tool. The time T gsp_ois the time from the start of each finger's operation to the end of each finger's operation. The hand operation time calculation unit 17 adds a predicted value of the time required to transmit a signal, which is the hand operation command, at time T. gsp_o , to reduce manual operation time T gsp to calculate. Over time T gsp_o A time that has elapsed before a difference between a simulated finger position and a target finger position becomes less than or equal to a predefined value can be taken into account.

[0038] So far, the example has been described in which the process time prediction unit 11 predicts the robot process time T1 and the manual process time T. gspUsing information about the object's position P and the values ​​of the insertion extent d and opening width w in the manual operation information, the operation time prediction unit 11 can predict the robot operation time T1 and the manual operation time T1. gsp Predicting the robot operation time T1 and the manual operation time T2 is achieved by using only a portion of the information about the object's position P and the values ​​of the insertion extent d and the opening width w. The operation time prediction unit 11 can use the value of the insertion extent d in predicting the robot operation time T1 and the manual operation time T2. gsp Do not use.

[0039] The preceding description is an example of a case in which the position of each finger of hand 31 is fixed immediately before grasping near the object, and hand 31 is moved at low speed in a linear direction and positioned. In this case, grip stability can be improved; for example, it is possible to easily grasp near the object's center of gravity by adjusting the insertion depth d. On the other hand, in a case where no straight path is provided near the object and hand 31 is positioned by high-speed movement along a curved path, the insertion depth d is used to predict the robot operation time T1 and the hand operation time T2. gsp unnecessary. Furthermore, in the case of a hand adsorbing the object, the value of the aperture width w is not relevant for predicting the hand action time T. gspunnecessary. In the case of the hand adsorbing the object, the hand action time T is gsp a predicted value of the time required to transmit a signal.

[0040] Fig. Figure 4 is a representation to explain an instruction to start the process of the hand 31 and an instruction to start the process of the robot 30 by the robot control device 10 according to the first embodiment. Fig. Figure 4 illustrates a diagram representing a change in the position of hand 31 and a diagram representing an open / closed state of hand 31. In the diagram representing the change in position of hand 31, the vertical axis represents the position of hand 31 and the horizontal axis represents a time T. In the diagram representing the open / closed state of hand 31, the vertical axis represents the open state or the closed state and the horizontal axis represents the time T.

[0041] The process time prediction unit 11 detects a time at which the value of the remaining robot process time T0 reaches a value that T0 <T gsp based on the value of the remaining robot operation time T0 and the value of the manual operation time T gsp fulfilled. Upon detecting the time, the process time prediction unit 11 transmits the reaching of the time when T0. <T gspOnce fulfilled, the manual operation start instruction unit 13 sends the value of the remaining robot operation time T0 at that time and the value of the manual operation time T. gsp at that time to the manual process start instruction unit 13.

[0042] Upon reaching the point in time when T0 <T gsp Once the condition is met, the hand operation start instruction unit 13 instructs the hand control device to start the process of hand 31 grasping the object, i.e., the closing process. The hand operation start instruction unit 13 sends a closing process command to the hand control device, thereby instructing the hand control device to start the closing process. By sending the hand operation command to hand 31 according to the instruction from the hand operation start instruction unit 13, hand 31 starts the closing process.

[0043] The manual operation start instruction unit 13 sends information specifying the time Tk, for example, time information specifying the time Tk, to the follow-up operation instruction unit 14. The time Tk is a time at which T0 <T gsp is fulfilled, and a time at which the manual operation start instruction unit 13 instructs the manual control device to start the closing process. Furthermore, the manual operation start instruction unit 13 sends T to the re-reaching operation instruction unit 14 as needed. gsp -Tp, where this is a difference between the value of the manual operation time T gsp and a value of an elapsed time Tp.

[0044] The process time prediction unit 11 can determine the extent of the correction ΔT and detect a time at which T0<(T gsp -ΔT) is satisfied, instead of the time at which T0 <T gspis fulfilled. When detecting the time, the process time prediction unit 11 transmits the reaching of the time when T0<(T gsp -AT) is fulfilled, to the manual operation start instruction unit 13. Additionally, the operation time prediction unit 11 sends the value of the remaining robot operation time T0 at the time, the value of the manual operation time T gsp at that time and a value of the magnitude of the correction ΔT to the manual process start instruction unit 13.

[0045] Here, an example of the magnitude of the correction ΔT is described. Here, the outer shape of the object is approximated by a sphere with a radius R. wapproximated. Additionally, the object is slightly deformed by receiving an external force. The opening width of hand 31 in the open state is denoted by w, and the opening width of hand 31 in a state where hand 31 is closed to grasp the object is denoted by wc. If wa, which is an opening width in the process of closing hand 31 from the open state, is 2R w When hand 31 comes into contact with the object.

[0046] Assume that hand 31 moves further out of the state of wa=2R w closes and the object is thereby deformed to such an extent that the width of the object in a direction in which the fingers of hand 31 are facing each other becomes 2R wc will be smaller than 2R wAt this point, a normal force Fn is generated between the object and the fingers of hand 31. Hand 31 can grasp the object by generating a frictional force against the normal force Fn and by geometrically constraining the object between its fingers. As described above, hand 31 continues to close to the extent that the opening width becomes 2R. wc This occurs after hand 31 begins to close and the opening width becomes 2R. w This results in a state where the fingers touch the object. In this case, the robot control device 10 instructs the hand control device to start the closing process, so that the opening width of the hand 31 is 2R. wcThis occurs when T0=0. Consequently, if ΔT>0, the robot control device 10 can bring the fingers into contact with the object at a time when the robot 30's operation is ending. Additionally, the robot control device 10 can reduce gripping failure by preventing the fingers from touching the object during the robot 30's operation.

[0047] It should be noted that in a case where the closing process is started when T0≤T is first detected gsp is fulfilled wa<2R wduring the robot 30's operation. In this case, the grasping may fail because the fingers of hand 31 begin to touch the object during the robot 30's operation. On the other hand, by appropriately setting the correction factor ΔT, the time at which the fingers begin to touch the object can coincide with the time at which the robot 30's operation ends. That is, by appropriately setting the correction factor ΔT, the opening width can be adjusted when T0=0, 2R w The success rate of grasping with hand 31 is therefore possible.

[0048] In the preceding description, the object's outer shape approximates a sphere, but the object's outer shape may approximate a rectangular parallelepiped. In this case, the success rate of grasping by hand 31 can be improved by appropriately determining the extent of the correction ΔT based on a length Lw of the object in the direction in which the fingers are opposite each other.

[0049] The adjustment time prediction unit 12 contains a function for calculating the adjustment time T. fit Based on the opening width w and at least one of the hand property information and the object property information, the adjustment time prediction unit 12 calculates the adjustment time T. fitby calculation, which integrates at least one of the stiffness Kh value or the viscosity Dh value as the hand property information and the stiffness Kw value or the viscosity Dw value as the object property information. The adaptation time prediction unit 12 sends the value of the adaptation time T. fit to the re-reaching process instruction unit 14.

[0050] In a case where any deformation of the object is insufficient, an inertial force is generated in the object due to the action of the robot 30, and a section of the object slides into contact with the fingers, causing the object to fall from the hand 31. By integrating at least one of the pieces of information regarding the physical properties of the finger contact sections of the fingers of hand 31 and the information regarding the deformation of the object into the calculation, the adaptation time prediction unit 12 can determine the adaptation time T.fit , with which the object can be sufficiently deformed. Consequently, it is possible to improve the success rate of grasping by hand 31. It should be noted that in the first embodiment, the deformation of the object includes a case in which the entire object is deformed and a case in which part of the object is deformed. The deformation of the object also includes a case in which only a protrusion or the like formed on the surface of the object is deformed. How the object is deformed is not limited to a specific appearance.

[0051] In the first embodiment, the function is for calculating the adjustment time T. fit a tabular database that establishes a relationship between the opening width w, the stiffness Kw of the object and the adjustment time T fitThe tabular database is not limited to a two-dimensional table of the opening width w and the stiffness Kw and can also contain a table of the opening width w and the adjustment time T. fit This applies to any type of object. The relationship between the opening width w, the stiffness Kw of the object, and the adjustment time T. fit This is obtained in advance by verifying the gripping process by the robot 30 and the hand 31. The creation of a table is described below, using as an example a table of the opening width w and the adjustment time T. fit is used for every type of object.

[0052] When collecting data to create a table, a large number of objects of varying sizes are prepared. For each of these objects, the success / failure of grasping is determined by verifying a series of operations in which hand 31 grasps the object and robot 30 lifts the object. Specifically, robot 30 is instructed to perform an operation in which lifting the object is initiated when a time T has elapsed. gsp +Tb elapses since the hand started the closing process, and the grasping success rate is calculated. A time Tb is incremented from 0 seconds in a predetermined step time. Verification is performed a predetermined number of times for each of the times T. gsp +Tb were performed where Tbs differ from each other, and the success rate is determined for each of the times of T. gsp+Tb is calculated based on the number of successful grasping attempts. Robot 30 is stopped in a state where the object is lifted, and whether the object is grasped at the position where robot 30 is stopped is determined using the image sensor.

[0053] Instead of the image sensor, a force sensor can be used to determine whether the object is gripped. Success / failure of the grip is determined based on an output from the force sensor. The force sensor is attached to a wrist section of the robot 30, which is a connecting part to the hand 31. Alternatively, a laser displacement sensor can be used instead of the image sensor, and success / failure of the grip can be determined based on a measurement result from the laser displacement sensor.

[0054] In a case where the calculated success rate is greater than or equal to a predetermined threshold, or the time Tb exceeds a predetermined upper limit Tb lim Once the time Tb is reached, the increase in time Tb is terminated in each step. The values ​​of the opening width w and the adjustment time T are entered into the table to be generated. fit When the increase in time Tb is complete, the values ​​are written together. The opening width w is taken as a value corresponding to the width of the object used for verification. The adjustment time T is written as a value. fit A value of time Tb is adopted in a case where the success rate is greater than or equal to the threshold.

[0055] In a case where the table contains the opening width w, the stiffness Kw and the adaptation time T fit The stiffness value Kw of the object is generated based on the gripping force value F. gspderived when hand 31 inserts the object with an opening width w, which is set to a predetermined width w0. The value of the gripping force F gsp The data is recorded based on the manual operation information entered into the robot control device 10. The table lists the values ​​for stiffness Kw, opening width w, and adaptation time T. fit written in conjunction with each other for each value of stiffness Kw.

[0056] In a case where the opening width w is an intermediate value between the values ​​shown in the table, that is, for the object whose width is the intermediate value, the adjustment time T fit by linear interpolation of a large number of values ​​of the adjustment time T fit calculated, as shown in the table. Values ​​used for linear interpolation in this case are, for example, values ​​of the adjustment time T. fit, which correspond to the respective values ​​of the opening width w, where these are adjacent data points on both sides of the intermediate value. Even in a case where the stiffness Kw of the object is an intermediate value between the values ​​shown in the table, the adjustment time T fit for the object by linear interpolation of a multitude of values ​​of the adjustment time T fit The values ​​shown in the table are calculated. For example, values ​​used for linear interpolation in this case are the values ​​of the adjustment time T. fit , which correspond to the respective values ​​of stiffness Kw, where the data points are adjacent on both sides of the intermediate value.

[0057] The value of the adjustment time T fitis entered into the re-reaching operation instruction unit 14. Information specifying the time Tk at which the instruction to start the gripping operation is given is entered from the manual operation start instruction unit 13 into the re-reaching operation instruction unit 14. Furthermore, T gsp -Tp, which represents a difference between the value of the manual operation time T gsp and the value of the elapsed time Tp, is entered as needed from the manual operation start instruction unit 13 into the retrieval operation instruction unit 14. The retrieval operation instruction unit 14 compares an elapsed time since T gsp -Tp 0 has been reached, with the adjustment time T fit The follow-up action instruction unit 14 indicates at the time when the elapsed time since T gsp -Tp 0 has been reached, the adjustment time T fitIf the predicted adjustment time T is exceeded, the command generation unit is instructed to issue the next robot operation command after the object has been grasped. In the manner described above, the re-grasping operation instruction unit 14 issues the command after the predicted adjustment time T has elapsed. fit Since hand 31 began grasping the object, an instruction is given to start the next operation of robot 30, which is to be carried out after hand 31 has completed the grasping operation. The drive unit propels robot 30 according to the robot operation command, and thus robot 30 starts the next operation to be carried out after grasping the object.

[0058] In a case where the robot operation time calculation unit 16 calculates the time T1c based on the robot operation command, the robot control device 10 can perform a process described below. In the preceding description, the robot operation time calculation unit 16 calculates the time T1c in the case where the robot 30 performs an operation according to the robot operation command. The time T1c may not be an operation time in the case where the robot 30 performs an operation according to the robot operation command, but it may be a total of operation times in a case where the robot 30 performs two consecutive operations according to the robot operation command. Here, in a case where the operation of moving hand 31 from point Ps to point Pe corresponds to one operation, the two operations correspond to the one operation and a subsequent operation of moving hand 31 from point Pe to point Pe2.The hand 31 is moved by the two processes at two consecutive positions Ps-Pe and Pe-Pe2.

[0059] For example, suppose that, according to the robot operation command, the two operations move hand 31 from point Ps to point Pe near the object and then cause hand 31 to perform a linear movement corresponding to the insertion depth d from point Pe to point Pe2. In this case, the robot operation time calculation unit 16 calculates the time T1c, which is the time from the start of hand 31's movement from point Ps to the end of hand 31's movement to point Pe2. The hand operation start instruction unit 13 sends the closing operation command to the hand control device at time Tk, where T0 <T gsp is fulfilled, regardless of whether before the start of the linear motion corresponding to the extent of the insertion d, or in the middle of the linear motion.

[0060] Next, a process procedure of the robot control device 10 is described. Fig. Figure 5 is a flowchart illustrating a process procedure of the robot control device 10 according to the first embodiment.

[0061] In step S1, the robot control device 10 acquires information about the object's position P and the manual operation information. In step S2, the robot control device 10 calculates the manual operation time T using the manual operation time calculation unit 17. gsp through a calculation that integrates the hand property information. Step S2 is a step for predicting the hand operation time T. gsp .

[0062] In step S3, the robot control device 10 calculates the adaptation time T. fit Step S3 of the adjustment time prediction unit is a step towards predicting the adjustment time T. fit The adjustment time prediction unit 12 calculates the adjustment time T.fitby calculation that integrates the hand operation information and at least one of the hand property information and the object property information. In step S4, the robot control device 10 calculates the remaining robot operation time T0 using the robot operation time calculation unit 16. Before calculating the remaining robot operation time T0, the operation time prediction unit 11 predicts the robot operation time T1 by performing a calculation that integrates the robot control system information, the position P information, and the hand operation information. The robot operation time calculation unit 16 calculates the remaining robot operation time T0 by subtracting the elapsed time Ta from the robot operation time T1. It should be noted that steps S2, S3, and a robot operation time prediction step T1 can be performed in any order.The robot control device 10 can simultaneously perform two or more of the procedures from step S2, step S3 and the step for predicting the robot operation time T1.

[0063] In step S5, the robot control device 10 determines, via the process time prediction unit 11, whether T0 <T GSP is fulfilled. If T0 <T gsp If the condition is not met (step S5, No), the robot control device 10 returns to step S4 and recalculates the remaining robot operation time T0. <T gsp If the condition is met (step S5, Yes), the robot control device 10 advances the procedure to step S6.

[0064] In step S6, the robot control device 10, via the hand operation start instruction unit 13, instructs the hand control device to start the operation of hand 31. Step S6 is a step to issue an instruction to start the operation of hand 31 at the time determined based on the predicted robot operation time T1 and the predicted hand operation time T1. gsp is determined.

[0065] In step S7, the robot control device 10 determines, via the re-reaching process instruction unit 14, whether the adaptation time T fit The time since the start of the grasping action has elapsed. If the adjustment time T fit If no time has elapsed since the start of the gripping process (step S7, No), the robot control device 10 repeats the procedure from step S7. If the adaptation time T fitSince the start of the grasping process (step S7, Yes), the robot control device 10 advances the procedure to step S8. In step S8, the robot control device 10, through the re-grasping operation instruction unit 14, issues an instruction for an operation of the robot 30 to be performed after the object grasping process. Step S8 is a step for issuing an instruction for the operation of the robot 30 to be performed after the object grasping process, once the predicted adjustment time T has elapsed. fit Since hand 31 began grasping the object, the robot control device 10 terminates the process of the Fig. 5 illustrated procedures.

[0066] According to the first embodiment, the robot control device 10 can, by predicting the robot operation time T1 through calculations that integrate the robot control system information, set a time for issuing the instruction to start the operation of hand 31, taking into account a delay in the operation of robot 30 due to the characteristics of the robot control system. By predicting the robot operation time T1 through calculations that integrate the hand operation information, the robot control device 10 can set the time for issuing the instruction to start the operation of hand 31, taking into account the operation of hand 31 as a function of the size of the object. By predicting the hand operation time T1 gspBy means of a calculation that integrates the hand property information, the robot control device 10 can set the time for issuing the instruction to start the operation of hand 31, taking into account the operation properties of hand 31. The robot control device 10 can precisely set the time for issuing the instruction for the operation of hand 31 in order to shorten the time required for the gripping operation, even if the setting is not carried out by a skilled person for an extended period. Thus, the robot control device 10 achieves the effect that, in order to shorten the time required for the gripping operation by robot 30 and hand 31, it is possible to precisely set the time for issuing the instruction for the operation of hand 31.

[0067] Furthermore, the robot control device 10, according to the first embodiment, specifies the adaptation time T. fitbeforehand and gives an instruction that the robot's process 30 after the gripping process after the predicted adjustment time T has elapsed fit since the hand 31 begins grasping the object. Even if trials to set a time to start the next grasping operation are not performed, the robot control device 10 can set the time to start the next grasping operation in a suitable manner.

[0068] A second embodiment is described in more detail below.

[0069] Fig. Figure 6 is a diagram illustrating an exemplary configuration of a robot control device 10A according to a second embodiment. The robot control device 10A includes a process time prediction unit 11A, which is similar to the process time prediction unit 11 described in the first embodiment. The robot control device 10A does not include the adaptation time prediction unit 12 described in the first embodiment. In the second embodiment, the same components as in the first embodiment are designated with the same reference numerals, and configurations are mainly described that differ from those of the first embodiment.

[0070] In the second embodiment, the re-reaching instruction unit 14 stores a waiting time value Tw, which is the time from when the hand 31 grasps the object until the robot 30 starts the next operation. The waiting time value Tw is determined in advance by a program for controlling the robot 30. Alternatively, the waiting time value Tw is determined in advance as a parameter value of the robot control device 10A. The waiting time Tw corresponds to a predefined adaptation time T. fit Different waiting times Tw can be set depending on the object. For example, in a case where a specific type of hand 31 is used, a time Tw1 can be set as the waiting time Tw in the case of grasping a metal object, and a time Tw2 can be set as the waiting time Tw in the case of grasping a resin object.

[0071] The re-reaching process instruction unit 14 compares an elapsed time since T gsp -Tp 0 has been reached, with the waiting time Tw. The follow-up action instruction unit 14 indicates at the time when the elapsed time since T gsp Once the robot reaches -Tp 0 and the waiting time Tw is exceeded, the instruction generation unit is instructed to issue the next robot action instruction after the object has been grasped. In the manner described above, after the waiting time Tw has elapsed since the hand 31 started grasping the object, the re-grasping action instruction unit 14 issues an instruction to start the robot 30's action to be performed after the object grasping process. The drive unit drives the robot 30 according to the robot action instruction, and the robot 30 thereby starts the next action to be performed after the object grasping process.

[0072] According to the second embodiment, the robot control device 10A, similar to the first embodiment, can precisely set a time for issuing a gripping instruction in order to shorten the time required for a gripping operation by the robot 30 and the hand 31. Even if trial and error is not performed to set a time for starting the next gripping operation, the robot control device 10A can also appropriately set the time for starting the next gripping operation.

[0073] A third embodiment is described in more detail below.

[0074] Fig. Figure 7 is a representation illustrating an exemplary configuration of a robot control device 10B according to a third embodiment. The robot control device 10B includes a gripper control parameter update unit 18, which differs from the robot control device 10 according to the first embodiment. Additionally, the robot control device 10B includes an operation time prediction unit 11B, which differs from the operation time prediction unit 11 described in the first embodiment, and an adjustment time prediction unit 12B, which differs from the adjustment time prediction unit 12 described in the first embodiment.In the third embodiment, the same components as in the first or second embodiment are designated with the same reference numerals as therein, and mainly configurations are described that differ from those of the first or second embodiment.

[0075] Information about the object's position P, the handling information, and success / failure information are entered into the gripping control parameter update unit 18. The success / failure information indicates whether the object was successfully gripped. The gripping control parameter update unit 18 updates a gripping control parameter. The gripping control parameter is a parameter used to control the gripping process. Here, the handling time T is... gsp and the adjustment time T fitsuch grasping control parameters. The grasping control parameter update unit 18 acquires the success / failure information, which indicates the result of determining the success / failure of grasping by verifying the operations of the robot 30 and the hand 31. It is possible to use information indicating the result of determining the success / failure of grasping in a grasping operation, which is the previous grasping operation, as the success / failure information. It should be noted that in a case where the verification is performed multiple times using the same value as the value of each grasping control parameter, the success / failure information is a value indicating a success rate of grasping. A case where the verification is performed multiple times using the same value as the value of each grasping control parameter is described below as an example.

[0076] The gripper control parameter update unit 18 receives information about success / failure and the hand operation information from the previous gripping operation. The gripper control parameter update unit 18 outputs new hand operation information to the operation time prediction unit 11B. This new hand operation information is updated information about an operation of hand 31 during the next gripping operation. Additionally, the gripper control parameter update unit 18 outputs information about the object's position P to the operation time prediction unit 11B. Like the operation time prediction unit 11 of the first embodiment, the operation time prediction unit 11B includes the robot operation time calculation unit 16. The operation time prediction unit 11B does not include the hand operation time calculation unit 17 described in the first embodiment.

[0077] Next, a process of the robot control device 10B is described. The operations of the robot 30 and the hand 31 are verified while each gripping control parameter is repeatedly updated. In the third embodiment, the gripping control parameter update unit 18 determines the value of the hand operation time T. gsp based on the success / failure information gathered by performing the verification, while the manual operation time T is updated gsp This process is repeated. Additionally, the gripping control parameter update unit 18 determines the value of the adaptation time T. fit based on the success / failure information gathered by performing the verification, while the adjustment time T is being updated fit is repeated.

[0078] The gripping control parameter update unit 18 updates the hand operation time T. gsp , by increasing the manual operation time T gspsets by specifying a value for the manual operation time T gsp with a specific step size Tg. The gripper control parameter update unit 18 stores a maximum value T. gsp_ max of manual operation time T gsp and a minimum value T gsp_ min of manual process time T gsp , if the manual operation time T gsp is updated repeatedly.

[0079] Additionally, the gripper control parameter update unit 18 stores a value for the step size Tg. It should be noted that a method in which the value of the manual operation time T gsp during the update of the manual operation time T gsp The modification is not limited to what is described in the third embodiment and can be modified if necessary.

[0080] The gripping control parameter update unit 18 updates the adaptation time T. fit by setting the adjustment time T fit, by assigning a value to the adjustment time T fit The gripper control parameter update unit 18 changes with a specific step size Tf. It stores a maximum value T. fit_ max of the adaptation time T fit and a minimum value T fit_ min of the adjustment time T fit , if the adjustment time T fit is updated repeatedly. Additionally, the gripper control parameter update unit 18 stores a value for the step size Tf. It should be noted that a method in which the value of the adaptation time T fit during the update of the adjustment time T fit The modification is not limited to what is described in the third embodiment and can be modified if necessary. Here, the process of the robot control device 10B is described when the value of the manual operation time T is gsp and the value of the adjustment time T fitThe gripping control parameter update unit 18 determines the value of the hand operation time T. gsp and then determines the value of the adjustment time T fit The robot control device 10B verifies the operations of the robot 30 and the hand 31 by setting the value of the adaptation time T. fit to the maximum value T fit_ maximum and sequential increase of the value of the manual operation time T gsp from the minimum value T gsp _min in steps of step size Tg. The robot control device 10B determines the value of the manual operation time T. gsp through such verification.

[0081] When starting the process to determine the value of the manual process time T gsp The gripping control parameter update unit 18 sets the minimum value T gsp _min as the value of the manual operation time T gsp and sets the maximum value T fit_ max as the value of the adjustment time T fitfixed. The gripping control parameter update unit 18 outputs the value of the hand operation time T. gsp , that is, the minimum value T gsp_ min, to the process time prediction unit 11B. The gripper control parameter update unit 18 outputs the value of the adjustment time T. fit , that is, the maximum value T fit_ max, to the adaptation time prediction unit 12B.

[0082] Similar to the first embodiment, the robot operation time prediction unit 11B calculates the robot operation time T1 by means of a calculation that integrates the robot control system information, the information about the position P, and the manual operation information, performed by the robot operation time calculation unit 16. The robot operation time prediction unit 11 predicts the robot operation time T1 by means of a calculation that integrates the robot control system information, the information about the position P, and the manual operation information. The robot operation time calculation unit 16 calculates the remaining robot operation time T0 by subtracting the time Ta, which is the elapsed time, from the robot operation time T1.

[0083] The process time prediction unit 11B compares a value of the remaining robot process time T0 with a value of the entered manual process time T. gspSimilar to the case of the first embodiment, the process time prediction unit 11B detects a time at which the value of the remaining robot process time T0 reaches a value that T0 <T gsp fulfilled. Upon detecting the time, the activity time prediction unit 11B transmits the reaching of the time when T0 <T gsp Once fulfilled, the manual operation start instruction unit 13 sends the value of the remaining robot operation time T0 at that time and the value of the manual operation time T. gsp at that time to the manual process start instruction unit 13.

[0084] Similar to the first embodiment, the manual operation start instruction unit 13 sends information specifying the time Tk, for example, time information specifying the time Tk, to the retrieval operation instruction unit 14. Furthermore, the manual operation start instruction unit 13 sends T to the retrieval operation instruction unit 14 as needed. gsp -Tp, where this is a difference between the value of the manual operation time T gsp , which was entered by the process time prediction unit 11B, and is the value of the elapsed time Tp.

[0085] Similar to the case of the first embodiment, the process time prediction unit 11B can determine the extent of the correction ΔT and detect the time at which T0 <T gsp -ΔT is satisfied, instead of the time at which T0 <T gsp This condition is met. Upon detecting the time, the process time prediction unit 11B transmits the reaching of the time when T0. <Tgsp -ΔT is satisfied, to the manual operation start instruction unit 13. Additionally, the operation time prediction unit 11B sends the value of the remaining robot operation time T0 at the time, the value of the manual operation time T gsp at that time and a value of the magnitude of the correction ΔT to the hand operation start instruction unit 13. Similar to the case of the first embodiment, the robot control device 10B instructs the hand control device to start the closing operation by the hand 31, such that the opening width of the hand 31 is 2R wc This occurs when T0=0. Consequently, if ΔT>0, the robot control device 10B can bring the fingers into contact with the object at the point in time when the robot 30's operation ends. Additionally, the robot control device 10B can reduce gripping failure by preventing the fingers from touching the object during the robot 30's operation.

[0086] The adjustment time prediction unit 12B gives the value of the adjustment time T. fit The information entered by the gripping control parameter update unit 18 is sent to the re-gripping operation instruction unit 14 as is. Similar to the first embodiment, the information specifying time Tk and T is entered into the re-gripping operation instruction unit 14. gsp -Tp, where the difference is between the value of the manual operation time T gsp and the value of the elapsed time Tp. The follow-up action instruction unit 14 indicates at the time when the elapsed time since T gsp -Tp 0 has been reached, the adjustment time T fitIf the threshold is exceeded, the command generation unit is instructed to issue the next robot operation command after grasping the object. The drive unit then propels robot 30 according to the robot operation command, and as a result, robot 30 starts the next operation to be performed after grasping the object.

[0087] After the robot 30's operation, the image sensor is used to determine whether the object is grasped at the position where the robot 30 stopped. As described above, the success / failure of the grasp is determined by verifying a series of operations in which hand 31 grasps the object and robot 30 lifts the object. The grasp success rate is calculated by causing robot 30 and hand 31 to perform operations a predetermined number of times. Gripping information specifying a success rate value is input into the grasp control parameter update unit 18. In a case where the success rate is greater than or equal to a predetermined threshold, the grasp control parameter update unit 18 outputs a value to the operation time prediction unit 11B, calculated by adding the step size Tg to the minimum value T. gsp _min of manual process time T gspThe gripping control parameter update unit 18 adds the step size Tg to the value of the hand operation time T. gsp , in order to reduce manual operation time T gsp to update. The robot control device 10B repeats operations similar to those described above with respect to the updated manual operation time T. gsp are described in order to capture the value of the success rate.

[0088] The gripping control parameter update unit 18 repeats the update of the hand operation time T. gsp and records the success rate up to the point at which the success rate becomes less than or equal to the threshold or at which the value of the manual operation time T gsp the maximum value T gsp _max is reached. In a case where the success rate is not less than or equal to the threshold and the manual process time T gsp the maximum value T gsp_max is reached while the manual operation time T is being updated and verified. gsp The gripping control parameter update unit 18 determines the maximum value T, which is repeated. gsp_ max as the value of the manual process time T gsp .

[0089] In a case where the success rate becomes less than or equal to the threshold, while the manual process time T is being updated and verified gsp The gripping control parameter update unit 18 determines the value of the hand operation time T, which is repeated. gsp the value of the manual processing time T gsp , which is determined in the verification process immediately before the verification where the success rate becomes less than or equal to the threshold.

[0090] In the manner described above, the gripping control parameter update unit 18 determines the value of the hand operation time T. gspbased on the success / failure information gathered by performing the verification, while the manual operation time T is updated gsp The process time prediction unit 11B provides a value for the predicted manual process time T. gsp , the value of the manual operation time T gsp from which is determined by the gripping control parameter update unit 18 based on the information about success / failure.

[0091] Next, the robot control device 10B performs a process to determine the value of the adaptation time T. fit The robot control device 10B verifies the operations of the robot 30 and the hand 31 by setting the value of the hand operation time T. gsp on the specific value and sequential reduction of the value of the adjustment time T fit from the maximum value T fit_The robot control device 10B determines the value of the adaptation time T in increments of step size Tf. fit through such verification.

[0092] The gripping control parameter update unit 18 sets the maximum value T fit_ max than the adjustment time T fit at the start of the process to determine the value of the adjustment time T fit fixed. The gripping control parameter update unit 18 outputs the value of the adaptation time T. fit , that is, the maximum value T fit_ max, to the adaptation time prediction unit 12B. Similar to the case of determining the value of the manual operation time T. gsp , also in the case of determining the value of the adjustment time T fitThe grasping success rate is calculated by having the robot (30) and the hand (31) perform operations a predetermined number of times. Grasping information specifying a success rate value is input to the grasping control parameter update unit (18). If the success rate is greater than or equal to the predetermined threshold, the grasping control parameter update unit (18) outputs a value to the adaptation time prediction unit (12B) calculated by subtracting the step size (Tf) from the maximum value (T). fit_ max of the adaptation time T fit The gripping control parameter update unit 18 updates the adaptation time T. fit by subtracting the step size Tf from the value of the adjustment time T fit The robot control device 10B repeats operations that correspond to the above with respect to the updated adaptation time T. fitThe described methods are similar in order to capture a calculated result of the success rate.

[0093] The gripping control parameter update unit 18 repeats the update of the adaptation time T. fit and records the success rate up to the point at which the success rate becomes less than or equal to the threshold or the value of the adjustment time T fit the minimum value T fit_ The minimum is reached. In a case where the success rate is not less than or equal to the threshold and the value of the adjustment time T is reached. fit the minimum value T fit_ min is reached while the update and verification of the adjustment time T fit The gripping control parameter update unit 18 determines the minimum value T, which is repeated. fit_ min as the value of the adjustment time T fitIn a case where the success rate becomes less than or equal to the threshold during the update and verification of the adjustment time T fit The gripping control parameter update unit 18 determines the value of the adaptation time T, which is repeated. fit the value of the adjustment time T fit , which is determined during the verification process immediately before the verification where the success rate becomes less than or equal to the threshold.

[0094] In the manner described above, the gripper control parameter update unit 18 determines the value of the adaptation time T. fit based on the success / failure information gathered by performing the verification, while the adjustment time T is being updated fit The process is repeated. The adjustment time prediction unit 12B provides a value for the predicted adjustment time T. fit the value of the adjustment time T fitfrom which the gripping control parameter update unit 18 determines the success / failure information. So far, the example described is where the gripping control parameter update unit 18 determines the manual operation time T. gsp and the adjustment time T fit updated and the value of the manual operation time T gsp and the value of the adjustment time T fit determined based on the information about success / failure. In the third embodiment, the gripping control parameter update unit 18 can determine the delay time T. rs of the robot's process 30 instead of the manual process time T gsp update and a value for the delay time T rs based on the information about success / failure. In the third embodiment, the gripping control parameters can determine the delay time T. rs , which is the robot control system information, and the adaptation time T fitbe.

[0095] The gripping control parameter update unit 18 determines the value of the delay time T. rs based on the success / failure information gathered by performing the verification, while the delay time T is updated rs The process is repeated. The process time prediction unit 11B calculates the robot process time T1 using the robot process time calculation unit 16 by calculating the value of the determined delay time T. rs Integrated. The process time prediction unit 11B predicts the robot process time T1 by calculation, which uses the value of the determined delay time T. rs integrated.

[0096] The gripping control parameter update unit 18 updates the delay time T rs by setting the delay time T rs by changing a value of the delay time T rswith a specific step size Tr. The gripper control parameter update unit 18 stores a maximum value T. rs_ max of the delay time T rs and a minimum value T rs_ min of the delay time T rs , if the delay time T rs is updated repeatedly.

[0097] Additionally, the gripper control parameter update unit 18 stores a value for the step size Tr. It should be noted that a method in which the value of the delay time T rs during the update of the delay time T rs The modification is not limited to what is described in the third embodiment and can be modified if necessary. Here, the process of the robot control device 10B is described when the value of the delay time T rs The gripping control parameter update unit 18 determines the value of the delay time T. rsand then determines the value of the adjustment time T fit The robot control device 10B verifies the operations of the robot 30 and the hand 31 by setting the value of the adaptation time T. fit to a fixed value and changing the value of the delay time T rs from the maximum value T rs_ max or minimum value T rs_ min with step size Tr. The robot control device 10B determines the value of the delay time T. rs through such verification.

[0098] When starting the process to determine the value of the delay time T rs The gripping control parameter update unit 18 sets the maximum value T rs_ max or minimum value T rs_ min as the value of the delay time T rs fixed. The gripping control parameter update unit 18 outputs the value of the delay time T. rsto the process time prediction unit 11B. Similar to the case of the first embodiment, the robot process time calculation unit 16 calculates the time T1c based on the robot process command. The robot process time calculation unit 16 calculates the robot process time T1 by adding the delay time T rs at time T1c. The robot process time calculation unit 16 calculates the remaining robot process time T0 by subtracting the time Ta, which is an elapsed time since the start of the process by the robot 30, from the robot process time T1.

[0099] The gripping control parameter update unit 18 repeats the update of the delay time T. rs and records the success rate until the success rate becomes less than or equal to the threshold or until the value of the delay time T rs the maximum value T rs_ max or minimum value T rs_The minimum time is reached. In the manner described above, the gripping control parameter update unit 18 determines the value of the delay time T. rs based on the success / failure information gathered by performing the verification, while the delay time T is updated rs The process time prediction unit 11B predicts the robot process time T1 by calculation, which uses the value of the determined delay time T. rs integrated.

[0100] According to the third embodiment, the robot control device 10B updates each of the manual operation time T. gsp and the adjustment time T fit and determines the value of the manual operation time T gsp and the value of the adjustment time T fit based on the success / failure information gathered through verification. Alternatively, the robot control device 10B updates each of the delay times T. rs, which is the robot control system information, and the adaptation time T fit and determines the value of the delay time T rs and the value of the adjustment time T fit Based on the success / failure information gathered through verification, the robot control device 10B can therefore improve the grasping success rate and reduce the time required for the grasping process by the robot 30 and the hand 31.

[0101] In the third embodiment, the grip control parameter update unit 18 can update the correction magnitude ΔT and determine a correction magnitude value based on success / failure information. In this case, the grip control parameters include the correction magnitude ΔT. The grip control parameter update unit 18 determines the correction magnitude value based on the success / failure information acquired by performing the verification, while the correction magnitude update ΔT is repeated. The grip control parameter update unit 18 updates the correction magnitude ΔT by adjusting the correction magnitude ΔT by changing a correction magnitude value with a specified step size.The process time prediction unit 11B calculates the robot process time T1 using the robot process time calculation unit 16, by means of a calculation that includes the value of the determined correction magnitude ΔT. Even when determining the value of the correction magnitude ΔT based on success / failure information, the robot control device 10B can easily improve the gripping success rate compared to a manual adjustment procedure by the user and can reduce the time required for the gripping process by the robot 30 and the hand 31.

[0102] In the third embodiment, the grip control parameter update unit 18 can update the insertion extent d and determine a value for the insertion extent d based on success / failure information. In this case, the grip control parameters include the insertion extent d. The grip control parameter update unit 18 determines the value of the insertion extent d based on the success / failure information acquired by performing the verification, while the insertion extent update is repeated. The grip control parameter update unit 18 updates the insertion extent d by adjusting the insertion extent d by changing a value of the insertion extent d with a specified step size.The process time prediction unit 11B calculates the robot process time T1 using the robot process time calculation unit 16, by means of a calculation that includes the value of the determined insertion depth d. Even in a case where the value of the insertion depth d is determined based on success / failure information, the robot control device 10B can easily improve the grasping success rate compared to a manual adjustment procedure by the user and can reduce the time required for the grasping process by the robot 30 and the hand 31.

[0103] A fourth embodiment is described in more detail below.

[0104] A fourth embodiment describes a modification of how the gripping control parameters are updated in the third embodiment. A process of the robot control device 10B according to the fourth embodiment differs from that of the third embodiment in the method for setting the gripping control parameters. Here, the process of the robot control device 10B according to the fourth embodiment is described with reference to Fig. 7 described. In the fourth embodiment, the same components as those in the first to third embodiments described above are designated with the same reference numerals as therein, and mainly configurations are described that differ from those of the first to third embodiments. In the fourth embodiment, the gripping control parameters are similar to those of the third embodiment, including the hand operation time T. gsp and the adjustment time T fitor the delay time T rs , which involves the robot control system information and the adaptation time T fit The gripping control parameters can include the degree of correction ΔT or the degree of insertion d. A description is given here using as an example a case in which the gripping control parameters determine the hand operation time T. gsp and the adjustment time T fit are. In the third embodiment, the robot control device 10B updates the manual operation time T. gsp or updates the adjustment time T fit by setting a value that is different from the value of the manual process time T gsp and the value of the adjustment time T fit to a fixed value and changing the other one with the specified step size Tg or Tf.

[0105] In the fourth embodiment, the robot control device 10B searches for a combination of the value of the manual operation time T. gspand the value of the adjustment time T fit using a search method such as particle swarm optimization, Bayesian optimization or a genetic algorithm.

[0106] A function for evaluating the brevity of an operation time is used as the evaluation function for the search. The operation time is the time required for the gripping operation by robot 30 and hand 31, and is the time from when robot 30 begins moving to the target position until hand 31 finishes gripping the object. In the event of a gripping failure, a significant penalty is added to the evaluation result. The gripper control parameter update unit 18 uses the evaluation function to search for an optimal combination of the hand operation time value T. gsp and the value of the adjustment time T fitto search for a suitable object that can be successfully grasped and thus reduce the operation time. The grasping control parameter update unit 18 terminates the search when the number of times the search is performed reaches a predetermined number and outputs a combination of the value of the hand operation time T. gsp and the value of the adjustment time T fit with the minimal rating function in the searches performed so far.

[0107] It should be noted that in a case where the gripping control parameters include the delay time T rs and the adjustment time T fit are, the gripping control parameter update unit 18 uses the evaluation function to search for an optimal combination of the value of the delay time T rs and the value of the adjustment time T fit to search for methods that are successful in grasping and can shorten the process time.

[0108] According to the fourth embodiment, the robot control device 10B uses the evaluation function to determine the optimal combination of the value of the manual operation time T. gsp and the value of the adjustment time T fit to search for a combination that can be successfully grasped and shorten the process time. Alternatively, the robot control device 10B uses the evaluation function to search for the optimal combination of the delay time value T. rs and the value of the adjustment time T fit to search for methods that can improve the gripping success rate and reduce the processing time. Consequently, the robot control device 10B can improve the gripping success rate and reduce the time required for the gripping process by the robot 30 and the hand 31.

[0109] A fifth embodiment is described in more detail below.

[0110] Fig. Figure 8 is a diagram illustrating an exemplary configuration of a robot control device 10C according to a fifth embodiment. The robot control device 10C includes a process time prediction unit 11C, which is similar to the process time prediction unit 11B described in the third embodiment. The robot control device 10C does not include the adaptation time prediction unit 12B described in the third embodiment. In the fifth embodiment, the same components as those in the first to fourth embodiments described above are designated with the same reference numerals as therein, and mainly configurations are described that differ from those of the first to fourth embodiments.

[0111] The robot control device 10C includes a gripper control parameter update unit 18C, which differs from the gripper control parameter update unit 18 described in the third embodiment. In the fifth embodiment, the gripper control parameter is the handling time T. gsp or the delay time T rs .

[0112] Similar to the case of the third embodiment, the gripping control parameter update unit 18C updates the hand operation time T. gsp and records the information about success / failure. The gripping control parameter update unit 18C determines the value of the hand operation time T. gsp based on the success / failure information gathered by performing the verification, while the manual operation time T is updated gspis repeated. Similar to the case of the third embodiment, the gripping control parameter update unit 18C alternatively updates the delay time T. rs , which is robot control system information, and records success / failure information. The gripper control parameter update unit 18C determines the value of the delay time T. rs based on the success / failure information gathered by performing the verification, while the delay time T is updated rs is repeated. In the fifth embodiment, the gripping control parameter update unit 18C does not update the adaptation time T. fit or determines the value of the adjustment time T fit .

[0113] The process of the robot control device 10C for determining the value of the manual operation time T gsp or the value of the delay time T rsThis is similar to the case of the third embodiment. According to the fifth embodiment, the robot control device 10C can improve the grasping success rate and reduce the time required for the grasping process by the robot 30 and the hand 31. It should be noted that, similar to the case of the third or fourth embodiment, the grasping control parameters can include the degree of correction ΔT or the degree of insertion d.

[0114] A sixth embodiment is described in more detail below.

[0115] In a sixth embodiment, an example is described in which a value of a gripping control parameter is determined by machine learning. Fig. Figure 9 is a diagram illustrating an exemplary configuration of a robot control device 10D according to the sixth embodiment. The robot control device 10D includes a gripper control parameter learning unit 20. The robot control device 10D does not include the gripper control parameter update unit 18 described in the third embodiment. The robot control device 10D includes an operation time prediction unit 11D, which is similar to the operation time prediction unit 11B described in the third embodiment. The robot control device 10D includes an adjustment time prediction unit 12D, which is similar to the adjustment time prediction unit 12B described in the third embodiment.In the sixth embodiment, the same components as those in the first to fifth embodiments described above are designated with the same reference numerals as therein, and mainly configurations are described that differ from those of the first to fifth embodiments.

[0116] The gripping control parameter learning unit 20 includes a learning device 21, an inference device 22, and a storage unit 23 for learned models. The learning device 21 learns a relationship between the object's position P, the handling information, and the gripping control parameters. This relationship is one in which the gripping success rate is greater than or equal to a predefined threshold and the operation time, which is the time required for the gripping process by the robot 30 and the hand 31, is minimized. The gripping control parameters are at least one of the hand operation time T. gsp , the adjustment time T fit and the delay time T rs , where this refers to the robot control system information. An example is given here of a case in which a relationship exists between the object's position P, the handling information, and the handling time T. gsp, the adjustment time T fit and the delay time T rs The manual operation information includes a value for the insertion extent d and a value for the opening width w. The manual operation information can also include a value for the object's width instead of the opening width w.

[0117] The learning device 21 generates a learned model that specifies the relationship between the object's position P, the handling information, and the gripping control parameters. The learned model storage unit 23 stores the learned model. The inference device 22 uses the learned model to determine the values ​​of the handling time T. gsp , the adjustment time T fit and the delay time T rs to derive.

[0118] Fig. Figure 10 is a representation illustrating the learning device 21 and the storage unit 23 for learned models in the gripper control parameter learning unit 20, which is included in the robot control device 10D according to the sixth embodiment. The learning device 21 includes a data acquisition unit 24 and a model generation unit 25. The values ​​of the delay time T rs , the manual operation time T gsp and the adjustment time T fit , the information about position P, the manual process information, the success / failure information and the process time information are entered into data acquisition unit 24.

[0119] The success / failure information is obtained by verifying the gripping process when updating the delay time T values. rs , the manual operation time T gsp and the adjustment time T fitFor each combination of position P, insertion depth d, and opening width w, the following information is obtained. In the sixth embodiment, the success / failure information is a value indicating the grasping success rate. The operation time information is a value indicating the length of time from when the robot 30 begins moving to the target position until the hand 31 completes the object grasping process. The operation time is measured when the grasping process is verified.

[0120] The data acquisition unit 24 generates training data using combinations of the delay time T values. rs , the manual operation time T gsp and the adjustment time T fitThe data includes information about position P, manual operation information, success / failure information, and operation time information. Data acquisition unit 24 extracts the delay time T from the combinations of input values. rs , the manual operation time T gsp and the adjustment time T fit A combination where the success rate is greater than or equal to the threshold and the operation time is shortest. Data acquisition unit 24 generates training data containing information about position P, manual operation information, and the extracted combination of delay time values ​​T. rs , the manual operation time T gsp and the adjustment time T fit are associated with each other. The data acquisition unit 24 records the learning data in the manner described above.

[0121] Using the training data, the model generation unit 25 creates a trained model for deriving the values ​​of the delay time T. rs , the manual operation time T gsp and the adjustment time T fit based on position P and the manual process information. The memory unit 23 for learned models stores the generated learned model.

[0122] A well-known algorithm such as supervised learning, unsupervised learning, or reinforcement learning can be used as a learning algorithm by the model generation unit 25. As an example, a case is described in which a neural network is applied. The model generation unit 25 learns combinations of the values ​​of the delay time T. rs , the manual operation time T gsp and the adjustment time T fitthrough so-called supervised learning according to a neural network model. Here, supervised learning is a process in which sets of data, each containing an input and an output, are given to the learning device 21, thereby learning properties in the training data and deriving outputs from inputs. The training data includes an input and a marker, which is an output associated with the input. The information about the position P and the manual process information correspond to inputs, and the values ​​of the delay time T correspond to inputs. rs , the manual operation time T gsp and the adjustment time T fit correspond to markings.

[0123] Fig. Figure 11 is a diagram illustrating an exemplary configuration of a neural network used for machine learning in the sixth embodiment. The neural network includes an input layer containing a plurality of neurons, a hidden layer (an intermediate layer) containing a plurality of neurons, and an output layer containing a plurality of neurons. The intermediate layer can be one layer, two, or more. Each value from a plurality of inputs to the input layer is multiplied by a weight and fed into the intermediate layer. Each value from a plurality of inputs to the intermediate layer is multiplied by a weight and output by the output layer.An output result printed from the output layer changes according to the value of the weight used for multiplication in the input layer and the value of the weight used for multiplication in the intermediate layer.

[0124] The neural network sets the values ​​of the weights so that results output by the output layer, after the position P and manual operation information have been input to the input layer, approximate the values ​​of the delay time Ts, the manual operation time Tgsp, and the adjustment time Tfit, thus generating combinations of the delay time T values. rs , the manual operation time T gsp and the adjustment time T fitThe model generation unit 25 creates a learned model by performing learning as described above. The model generation unit 25 can read a learned model that has already been created from the learned model storage unit 23 and update the learned model by relearning according to the training data. Fig. Figure 12 is a representation illustrating the inference device 22 and the storage unit 23 for learned models in the gripper control parameter learning unit 20, which is included in the robot control device 10D according to the sixth embodiment. The inference device 22 includes a data acquisition unit 26 and an inference unit 27.

[0125] The information about position P and the manual operation information is entered into the data acquisition unit 26, which then acquires the information about position P and the manual operation information, which constitutes inference data. The inference unit 27 reads a learned model from the storage unit 23 for learned models. The inference unit 27 inputs the information about position P and the manual operation information into the learned model, thereby determining the values ​​of the delay time T. rs , the manual operation time T gsp and the adjustment time T fit will be issued.

[0126] The gripping control parameter learning unit 20 outputs the values ​​of the delay time T. rs and the manual operation time T gsp to the process time prediction unit 11D. The gripper control parameter learning unit 20 outputs the value of the adaptation time T. fitto the 12D adaptation time prediction unit.

[0127] Similar to the first embodiment, the robot operation time calculation unit 16 calculates the time T1c based on the robot operation command. The robot operation time calculation unit 16 calculates the robot operation time T1 by adding the delay time T. rsat time T1c. The robot operation time calculation unit 16 calculates the remaining robot operation time T0 by subtracting the time Ta, which is the elapsed time since the robot 30 started the operation, from the robot operation time T1. Similar to the first embodiment, the operation time prediction unit 11D calculates the robot operation time T1 by means of a calculation that integrates the robot control system information, the information about the object's position P, and the hand operation information. The operation time prediction unit 11D predicts the robot operation time T1 by means of a calculation that integrates the robot control system information, the information about the object's position P, and the hand operation information.

[0128] The process time prediction unit 11D compares the value of the remaining robot process time T0 with the value of the entered manual process time T.gsp . Similar to the case of the first embodiment, when the time is detected at which the value of the remaining robot operation time T0 reaches a value that T0< Tgsp Once fulfilled, the process time prediction unit 11D transmits the reaching of the time T0. <T gsp fulfilled, to the manual operation start instruction unit 13. Additionally, the operation time prediction unit 11D sends the value of the remaining robot operation time T0 at that time and the value of the manual operation time T. gsp at that time to the manual process start instruction unit 13.

[0129] It should be noted that, similar to the case of the first embodiment, the process time prediction unit 11D can determine the extent of the correction ΔT and can detect the time at which T0 <T gsp -ΔT is satisfied, instead of the time at which T0 <T gspThe robot control device 10D instructs the hand control device to start the gripping process, so that the opening width of the hand is 31 to 2R. wc This occurs when T0=0. Consequently, if ΔT>0, the robot control device 10D can bring the fingers into contact with the object at the point in time when the robot 30's operation ends. Additionally, the robot control device 10D can reduce gripping failure by preventing the fingers from touching the object during the robot 30's operation.

[0130] The 12D adjustment time prediction unit gives the value of the adjustment time T. fit The input from the gripping control parameter learning unit 20 is sent to the re-gripping process instruction unit 14 as is. It should be noted that the process from each of the manual process start instruction unit 13 and the re-gripping process instruction unit 14 is similar to that in the case of the first embodiment.

[0131] The sixth embodiment describes the case in which supervised learning is applied to the learning algorithm used by the model generation unit 25, but learning other than supervised learning can also be applied to the learning algorithm. The model generation unit 25 can perform machine learning using a learning algorithm such as reinforcement learning, unsupervised learning, or semi-supervised learning. The model generation unit 25 can perform machine learning using a learning algorithm such as deep learning, genetic programming, inductive logic programming, or a support vector machine.

[0132] In the sixth embodiment, the learning device 21 is included in the robot control device 10D. The learning device 21 can be a device located outside the robot control device 10D. The learning device 21 can be a device connected to the robot control device 10D via a network, or a device located on a cloud server.

[0133] The learning device 21 is not limited to learning the values ​​of the gripper control parameters according to learning data generated for a single robot control device 10D. The learning device 21 can learn the values ​​of the gripper control parameters according to learning data generated for a multitude of robot control devices 10D. The learning device 21 can acquire learning data from a multitude of robot control devices 10D used at the same location, or it can acquire learning data from a multitude of robot control devices 10D used at different locations. The learning data can be acquired from the robot control devices 10D operating independently at a multitude of locations. After starting the acquisition of learning data from the multitude of robot control devices 10D, a new robot control device 10D can be added as a target from which the learning data will be acquired.Additionally, after the acquisition of learning data from the multitude of robot control devices 10D has been started, a portion of the multitude of robot control devices 10D can be excluded from the targets from which the learning data is being acquired.

[0134] The learning device 21, which has learned one robot control device 10D, can learn another robot control device 10D that is not the same robot control device 10D. The learning device 21 can update the learned model by relearning the other robot control device 10D.

[0135] The learning device 21 only needs to establish a relationship between the position P of the object, the handling information and at least one of the values ​​of the handling time T. gsp , the adjustment time T fit and the delay time T rslearning. Using the learned model, the inference device 22 derives at least one of the values ​​of the manual operation time T. gsp , the adjustment time T fit and the delay time T rs from the object's position P and the manual process information.

[0136] The manual operation information to be entered into the learning device 21 must include at least one of the values ​​of the insertion extent d and the opening width w. The manual operation information to be entered into the learning device 21 may include information about speed or acceleration instead of the insertion extent d if the robot 30 is caused to operate at a fixed position by the insertion extent d.

[0137] According to the sixth embodiment, the robot controller 10D learns the relationship between the object's position P, the hand operation information, and the gripping control parameters. This relationship is one in which the gripping success rate is greater than or equal to a predefined threshold, and the operation time—the time required for the gripping operation by the robot 30 and the hand 31—is minimized. The robot controller 10D uses the learned model to derive the values ​​of the gripping control parameters based on the object's position P and the hand operation information. Consequently, the robot controller 10D can improve the gripping success rate and reduce the time required for the gripping operation by the robot 30 and the hand 31.

[0138] A seventh embodiment is described in more detail below.

[0139] Fig. Figure 13 is a figure illustrating an exemplary configuration of a robot control device 10E according to a seventh embodiment. Similar to the robot control device 10D according to the sixth embodiment, the robot control device 10E includes the gripper control parameter learning unit 20. The robot control device 10E includes an operation time prediction unit 11E, which is similar to the operation time prediction unit 11D described in the sixth embodiment. The robot control device 10E does not include the adaptation time prediction unit 12D described in the sixth embodiment. In the seventh embodiment, the same components as those in the first to sixth embodiments described above are designated with the same reference numerals as therein, and mainly configurations are described that differ from those of the first to sixth embodiments.

[0140] In the seventh embodiment, the gripping control parameters are at least one dependent on the manual operation time T. gsp and the delay time T rs , where this refers to the robot control system information. The learning device 21 learns a relationship between the object's position P, the handling information, and at least one of the values ​​of the handling time T. gsp and the delay time T rs In the seventh embodiment, the value of the adaptation time T is fit not entered into the learning device 21. The inference device 22 uses the learned model to determine the values ​​of the manual operation time T. gsp and the delay time T rs to derive. The inference device 22 derives the value of the adaptation time T. fit not off.

[0141] Similar to the sixth embodiment, the robot controller 10E learns the relationship between the object's position P, the hand operation information, and the gripping control parameters. This relationship is one in which the gripping success rate is greater than or equal to a predetermined threshold, and the operation time—the time required for the gripping operation by the robot 30 and the hand 31—is minimized. As in the sixth embodiment, the robot controller 10E uses the learned model to derive values ​​for the gripping control parameters based on the position P and the hand operation information. Consequently, the robot controller 10E can improve the gripping success rate and reduce the time required for the gripping operation by the robot 30 and the hand 31.

[0142] Next, a hardware configuration is described that implements the robot control devices 10, 10A, 10B, 10C, 10D, and 10E according to the first through seventh embodiments. The robot control devices 10, 10A, 10B, 10C, 10D, and 10E are each implemented by a processing circuit. The processing circuit can be a circuit in which a processor executes software, or it can be a dedicated circuit.

[0143] In a case where the processing circuit is implemented by software, the processing circuit is, for example, one in Fig. 14 illustrated control circuits. Fig. Figure 14 is a diagram illustrating an exemplary configuration of a control circuit 50 according to the first to seventh embodiments. The control circuit 50 includes an input unit 51, a processor 52, a memory 53, and an output unit 54. The input unit 51 is an interface circuit that receives data input from outside the control circuit 50 and transmits the data to the processor 52. The output unit 54 is an interface circuit that sends data from the processor 52 or the memory 53 to an external location outside the control circuit 50. In a case where the processing circuit is configured to receive the input data, the output unit 54 is configured to receive the input data from the processor 52 or the memory 53. Fig. In the illustrated control circuit 50, processor 52 reads and executes a robot control program stored in memory 53, thereby implementing each component of the robot control devices 10, 10A, 10B, 10C, 10D, and 10E. The robot control program is a program corresponding to each component of the robot control devices 10, 10A, 10B, 10C, 10D, and 10E. Additionally, processor 52 outputs data, such as a calculation result, to a volatile memory of memory 53. Memory 53 is also used as temporary storage during each process performed by processor 52. Processor 52 can output and store data, such as a calculation result, in memory 53, or it can store data, such as a calculation result, in an auxiliary storage device via the volatile memory of memory 53.A function for storing information in each component is implemented by memory 53 or the auxiliary storage device.

[0144] The processor 52 is a central processing unit (CPU, also referred to as a central processing device, a processing device, an arithmetic device, a microprocessor, a microcomputer, a processor, or a digital signal processor (DSP), or the like). The memory 53 corresponds, for example, to non-volatile or volatile semiconductor memory, such as random access memory (RAM), read-only memory (ROM), flash memory, erasable programmable ROM (EPROM), or electrically erasable programmable PROM (EEPROM (registered trademark)), a magnetic disk, a flexible disk, an optical disk, a compact disk, a mini disk, or a digital versatile disk (DVD).

[0145] Fig. 14 is an example of hardware in a case where each component is implemented by the processor 52 and the memory 53 for general use, but each component can be implemented by a dedicated hardware circuit. Fig. Figure 15 is a representation illustrating an exemplary configuration of a hardware circuit 55 as a dedicated circuit according to the first to seventh embodiments.

[0146] The hardware circuit 55, as a dedicated circuit, includes an input unit 51, an output unit 54, and a processing circuit 56. The processing circuit 56 can be a standalone circuit, a compound circuit, a programmed processor, a parallel programmed processor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or a combination thereof. Each component can be implemented by combining the control circuit 50 and the hardware circuit 55.

[0147] It should be noted that the robot control program can be stored on a recording medium, such as a compact disc (CD)-ROM or a DVD-ROM, and the recording medium can be provided to implement any embodiment. List of reference symbols

[0148] 10, 10A, 10B, 10C, 10D, 10E Robot control device; 11, 11A, 11B, 11C, 11D, 11E Operation time prediction unit; 12, 12B, 12D Adjustment time prediction unit; 13 Manual operation start instruction unit; 14 Reaching action instruction unit; 15 Storage unit; 16 Robot operation time calculation unit; 17 Handling time calculation unit; 18, 18C Gripping control parameter update unit; 20 Gripping control parameter learning unit; 21 Learning device; 22 Inference device; 23 Memory unit for learned models; 24, 26 Data acquisition unit; 25 Model generation unit; 27 Inference unit; 30 Robot; 31 Hand; 40 Robot control system; 50 Control circuit; 51 Input unit; 52 Processor; 53 Memory; 54 Output unit; 55 Hardware circuit; 56 Processing circuit.

Claims

[1] Robot control device (10), comprising: a process time prediction unit (11) for predicting a robot process time (T1), which is a time required by a robot (30) to cause a hand (31) of the robot (30) to reach a target position (P) get ) achieved, and a manual operation time (T gsp ), where this is a time from the time at which an action of the hand (31) is commanded until the time at which the hand (31) carries out an action of grasping an object at the target position (P get ) finished, acts; and a hand operation start instruction unit (13) for issuing an instruction to start an operation of the hand (31) at a time determined based on the predicted robot operation time (T1) and the predicted hand operation time (T gsp ) is determined, whereby the process time prediction unit (11) predicts the robot process time (T1) by calculation, which integrates the robot control system information, which includes a control parameter to determine the responsiveness of a control system of the robot (30). [2] Robot control device (10E), comprising: a process time prediction unit (11E) for predicting a robot process time (T1), which is the time a robot (30) needs to cause a hand (31) of the robot (30) to move to a target position (P) get ) achieved, and a manual operation time (T gsp ), where this is a time from the time at which an action of the hand (31) is commanded until the time at which the hand (31) carries out an action of grasping an object at the target position (P get ) finished, acts; a hand operation start instruction unit (13) for issuing an instruction to start an operation of the hand (31) at a time determined based on the predicted robot operation time (T1) and the predicted hand operation time (T gsp ) is determined; and a grasp control parameter learning unit (20) for learning a relationship between information about a position of the object, hand process information, which is information about a process of the hand (31) and which are obtained based on information about the object, and a grasping control parameter, wherein the relationship is a relationship where a grasping success rate is greater than or equal to a predetermined threshold and an operation time, which is the time required for a grasping operation to grasp the object, is shortest, wherein the process time prediction unit (11E) predicts the robot process time (T1) by calculation, integrates the robot control system information that specifies a property of a control system of the robot (30), and the gripping control parameter at least one of the manual process time (T gsp ) and the robot control system information. [3] Robot control device (10), comprising: a process time prediction unit (11) for predicting a robot process time (T1), which is a time required by a robot (30) to cause a hand (31) of the robot (30) to reach a target position (P) get ) achieved, and a manual operation time (T gsp ), where this is a time from the time at which an action of the hand (31) is commanded until the time at which the hand (31) carries out an action of grasping an object at the target position (P get ) finished, acts; a hand operation start instruction unit (13) for issuing an instruction to start an operation of the hand (31) at a time determined based on the predicted robot operation time (T1) and the predicted hand operation time (T gsp ) is determined; an adjustment time prediction unit (12) for predicting an adjustment time, which is the time from when the hand (31) begins to grasp the object until when the hand (31) fits the object; and a re-grasping action instruction unit (14) for issuing an instruction for an action of the robot (30) to be carried out after an action of grasping the object after the predicted adjustment time has elapsed from the start of grasping the object by the hand (31), wherein The process time prediction unit (11) predicts the robot process time (T1) by calculation, integrating robot control system information that specifies a property of a control system of the robot (30). [4] Robot control device (10) according to claim 2 or 3, wherein the process time prediction unit (11) predicts the robot process time (T1) by calculation which integrates the robot control system information, information about a position of the object and hand process information which is information about a process of the hand (31) when grasping the object. [5] Robot control device (10) according to one of claims 1 to 4, wherein the process time prediction unit (11) predicts the manual process time (T gsp ) predicts by calculation, which integrates hand property information that specifies a property of a control system of the hand (31) or a property of a mechanism forming the hand (31). [6] Robot control device (10C) according to any one of claims 1 to 5, comprising: a gripping control parameter update unit (18C) for updating the hand operation time (T gsp ), which is a grasping control parameter for controlling a grasping process by the robot (30) and the hand (31), and for capturing success / failure information that indicates a result of determining success / failure of the grasping by verifying operations of the robot (30) and the hand (31), wherein the gripping control parameter update unit (18C) a value of the hand operation time (T gsp ) is determined based on the success / failure information gathered by performing the verification, while an update of the manual operation time (T gsp ) is repeated, and the process time prediction unit (11C) a value of the manual process time (T gsp), which is determined based on the information about success / failure, as a value of the predicted manual process time (T gsp ) outputs. [7] Robot control device (10C) according to any one of claims 1 to 5, comprising: a gripping control parameter update unit (18C) for updating the robot control system information, which is a gripping control parameter for controlling a gripping operation by the robot (30) and the hand (31), and for capturing success / failure information indicating a result of determining success / failure of the gripping by verifying operations of the robot (30) and the hand (31), wherein The gripper control parameter update unit (18C) determines the robot control system information based on the success / failure information acquired by performing the verification, while an update of the hand operation time (T gsp) is repeated, and The process time prediction unit (11C) predicts the robot process time (T1) by calculation, which integrates the robot control system information determined on the basis of the success / failure information. [8] Robot control device (10E) according to claim 1 or 3, comprising a gripping control parameter learning unit (20) for learning a relationship between information about a position of the object, hand operation information, which is information about an operation of the hand (31) and is obtained on the basis of information about the object, and a gripping control parameter, which is at least one of the hand operation time (T) gsp) and the robot control system information, where the relationship is one in which a grasping success rate is greater than or equal to a predetermined threshold and an operation time, which is the time required for a grasping operation to grasp the object, is shortest. [9] Robot control device (10) according to claim 1, comprising: an adjustment time prediction unit (12) for predicting an adjustment time, that is, a time from the time at which the hand (31) begins to grasp the object until the time at which the hand (31) fits the object; and a re-grasping action instruction unit (14) for issuing an instruction for an action of the robot (30) to be carried out after an action of grasping the object after an elapse of the adjustment time which is predicted from the start of grasping the object by the hand (31). [10] Robot control device (10) according to claim 9, wherein the adaptation time prediction unit (12) predicts the adaptation time by calculation, which integrates at least one of hand property information and object property information, wherein the hand property information specifies a property of a control system of the hand (31) or an operation property of the hand (31), wherein the object property information specifies a property of the object, and hand operation information, which is information about an operation of the hand (31) and which is obtained on the basis of information about the object. [11] Robot control device (10B) according to claim 9 or 10, comprising: a gripping control parameter update unit (18) for updating each of the manual operation time (T gsp) and the adjustment time, each of which is a grasping control parameter for controlling a grasping process by the robot (30) and the hand (31), and for capturing information about success / failure, which indicates a result of determining success / failure of grasping, by verifying operations of the robot (30) and the hand (31), wherein the gripping control parameter update unit (18) a value of the hand operation time (T gsp ) is determined based on the success / failure information gathered by performing the verification, while an update of the manual operation time (T gsp ) is repeated, and a value for the adjustment time is determined based on the success / failure information gathered by performing the verification, while the adjustment time update is repeated, the process time prediction unit (11B) a value of the manual process time (T gsp), which is determined based on the information about success / failure, as a value of the predicted manual process time (T gsp ) spends, and The adjustment time prediction unit (12B) outputs a value of the adjustment time, which is determined based on the information about success / failure, as a predicted adjustment time value. [12] Robot control device (10B) according to claim 9 or 10, comprising: a gripping control parameter update unit (18) for updating each of the robot control system information and the adaptation time, each of which is a gripping control parameter for controlling a gripping operation by the robot (30) and the hand (31), and for acquiring success / failure information indicating a result of determining success / failure of the gripping by verifying operations of the robot and the hand (31), wherein the gripper control parameter update unit (18) determines the robot control system information based on the success / failure information acquired by performing the verification, while the robot control system information update is repeated, and determines the adaptation time based on the success / failure information acquired by performing the verification, while the adaptation time update is repeated, The process time prediction unit (11B) predicts the robot process time (T1) by calculation, which integrates the robot control system information determined on the basis of the success / failure information, and The adjustment time prediction unit (12B) outputs a value of the adjustment time, which is determined based on the information about success / failure, as a predicted adjustment time value. [13] Robot control device (10D) according to claim 9 or 10, comprising a gripping control parameter learning unit (20) for learning a relationship between a position of the object, hand operation information, which is information about an operation of the hand (31) and is obtained on the basis of information about the object, and a gripping control parameter, which is at least one of the hand operation time (T) gsp ), the adaptation time and the robot control system information, where the relationship is one in which a grasping success rate is greater than or equal to a predetermined threshold, and an operation time, which is the time required for a grasping operation to grasp the object, is shortest. [14] Robot control methods, including: a step towards predicting a robot operation time (T1), that is, a time that a robot (30) needs to cause a hand (31) of the robot (30) to move to a target position (P) get ) achieved by calculation that integrates the robot control system information, which includes a control parameter for determining the responsiveness of a control system of the robot (30); a step towards predicting a manual operation time (T gsp ), that is, a time from the time at which an action of the hand (31) is commanded, until the time at which the hand (31) carries out an action of grasping an object at the target position (P get ) finished; and a step to issue an instruction to start a hand operation (31) at a time determined based on the predicted robot operation time (T1) and the predicted hand operation time (T gsp ) is determined. [15] Computer-readable storage medium (53) on which a robot control program is stored that causes a computer system to perform the following: a step towards predicting a robot operation time (T1), that is, a time that a robot (30) needs to cause a hand (31) of the robot (30) to move to a target position (P) get ) achieved by calculation that integrates the robot control system information, which includes a control parameter for determining the responsiveness of a control system of the robot (30); a step towards predicting a manual operation time (T gsp ), that is, a time from the time at which an action of the hand (31) is commanded, until the time at which the hand (31) carries out an action of grasping an object at the target position (P get ) finished; and a step to issue an instruction to start a hand operation (31) at a time determined based on the predicted robot operation time (T1) and the predicted hand operation time (T gsp ) is determined.

Images