JP5382359B2 - Robot system - Google Patents

Description

  In the present invention, for example, in fitting / inserting work (assembly work) of a gripping work by a robot arm and a hand, the work quality, that is, work success / work failure is determined with high accuracy using sensor information, and the work failure is determined. The present invention relates to a robot system capable of estimating the cause of work failure in the case of being performed.

  First, a general industrial robot will be described, and then the system configuration of impedance control using force sensor technology and fitting / insertion by impedance control in fitting / insertion work performed using such an industrial robot. A method for realizing the work will be described.

  Here, the impedance control is a position for setting the mechanical impedance (inertia, damping coefficient, rigidity) generated when an external force is applied to the hand of the robot to a value convenient for the intended work. It is a force control technique.

  The impedance control includes a passive impedance method in which a mechanical element such as a spring or a damper is attached to the hand of the robot to change the hand impedance, and feedback control using measured values of the hand position, speed, force, etc. There is an active impedance method for changing the impedance.

  FIG. 1 is a block diagram of an industrial robot according to the prior art. In FIG. 1, a robot 101 is a manipulator having a plurality of joint axes and links. Here, the “manipulator” means a machine that is composed of segments connected to each other and that is intended to grab (grab) or move an object (part, tool, etc.). Each joint shaft of the manipulator has a built-in drive motor with an encoder, and each joint can be driven independently. The controller 102 is a control unit of the robot 101 and is a device for controlling the motion of the robot 101 by configuring feedback control (position control system) based on the encoder signals of the respective joint axis drive motors. The portable teaching operation panel 103 is an interface for a teacher to manually (JOG) operate the robot and create / edit an operation program. The portable teaching operation panel 103 mainly includes an operation button group 103a and a display screen 103b. The end effector 104 is provided on the wrist portion of the robot 101. In the case of FIG. 1, a hand (gripper) for gripping a component is attached, but the present invention is not limited to this, and various tools can be attached to the end effector according to the application.

  FIG. 2 shows a block diagram of an industrial robot in the prior art with impedance control. The 6-axis force sensor 105 is attached to the wrist of the robot 101, and can measure the forces in the XYZ directions and the moments around the axes. The impedance control unit 106 is configured inside the controller 102 and configures a feedback control system based on the output of signals from the force sensor 105 and each axis encoder. The impedance control unit 106 outputs a torque command or a current command for each drive motor, and the actuator drive amplifier unit 107 supplies power to each drive motor based on the torque (or current) command value. The operation program storage unit 108 stores the operation program created (taught) on the portable teaching operation panel 103 in the controller. The operation program execution unit 109 interprets and executes the operation program stored in the operation program storage unit 108 and gives an operation command to the impedance control unit 106. The workpiece 110a is a fitting component that is gripped by the end effector 104, and the target workpiece 110b is a fitting component to be fitted and inserted. The workpiece fixing jig 111 fixes the target workpiece 110b. In general, a cylinder is driven by compressed air to fix a workpiece.

  With such a control system, by setting the robot 101 in the impedance control state and appropriately adjusting the parameters, it becomes possible to allow the parts to be fitted while allowing the position and orientation error.

  FIG. 3 shows a control block diagram for implementing impedance control in the prior art. In FIG. 3, the position control system 106a is based on the position command of each joint axis input from the operation program execution unit and the current position (feedback) input from the encoder, and the torque (or current command) of each drive motor. Is output to the actuator drive amplifier unit 107. Fref is the force moment command (force moment target value), and Ffb is the force moment feedback value. θref is a position command (joint coordinate system) sent from the operation program execution unit 109, and δθ is a position correction amount calculated by the impedance control calculation unit 106b. First, the impedance control calculation unit 106b calculates a position correction amount δP in the orthogonal coordinate system based on Fref and Ffb according to the following equation (106c).

δP = (Ms ² + Ds + K) ⁻¹ (Fref−Ffb) (1)
Here, M, D, and K are an inertia matrix, a viscosity coefficient matrix, and a stiffness matrix (spring constant), respectively. Usually, these are set as a diagonal matrix, and impedance characteristics independent of each axis are set. Moreover, s is a Laplace operator and corresponds to the first derivative with respect to time.
The position correction amount δP in the orthogonal coordinate system is decomposed into the position correction amount δθ in the joint coordinate system using the Jacobian matrix J (θ) by the following equation (106d).

δθ = J (θ) ^-1 δP (2)
By giving a position command θref ′ obtained by adding δθ to θref to the position control system 106a, the robot operates while maintaining the characteristics designated by M, D, and K with respect to external force and moment. For example, the robot operates like a spring against an external force due to K, and at that time, M and D are reduced to operate lightly and smoothly.

Next, an insertion operation by impedance control in the prior art will be described.
In order to perform the insertion work by impedance control, first, as shown in FIG. 4A, the fitting part 110a is gripped by the hand 104 of the robot 101 and aligned on the axis of the hole of the fitting part 110b. Here, as shown in FIG. 4A, the X axis is defined as shown so that the insertion direction of the fitting part 110a is the Z axis direction and the XY plane is perpendicular to the Z axis. Define the Y axis to be perpendicular to the axis and the Z axis. In this state, the robot 101 is switched from the position control state to the impedance control state, and an appropriate force command is given from the operation program execution unit 109 to the impedance control unit 106, whereby, as shown in FIG. The fitting part 110a can be inserted to the bottom of the fitting part 110b while absorbing the error and the rotation error about the X axis and the Y axis. At this time, in general, whether the movement (insertion amount) of the hand 104 from the contact between the workpieces to the bottom of the hole coincides with the depth L of the hole, whether the operation is good, that is, the operation is successful. It can be determined whether (insertion amount is sufficient) / work failure (insertion amount is insufficient). However, as shown in FIG. 4C, when the fitting part 110a is a thin cylindrical terminal connector and the fitting part 110b is a pin-shaped terminal connector, the parts shown in FIGS. 4D and 4E are shown. As described above, a situation in which the quality of work cannot be determined only by the amount of insertion may occur. That is, in FIGS. 4D and 4E, the movement amount of the hand 104 is both L, but it is determined that the work is successful in the case of FIG. 4D and the work is unsuccessful in the case of FIG. It should be.

  Patent Literature 1 and Patent Literature 2 are disclosed as techniques for determining work quality. In Patent Document 1, normality of a work execution state is obtained by collating data sampled from a sensor during work execution with a state monitoring table in which sensor data (allowable range) at the time of work success is recorded. -An abnormality is judged. In Patent Document 2, table (motion correction table) data for correcting robot motion based on data sampled from a sensor (working state) is automatically created.

JP-A-63-89283 JP-A-8-174456

  Since Patent Document 1 only holds a range of output values (minimum value and maximum value) of the force sensor and the tactile sensor at the time of execution of each step (for example, MOVE command) of the operation program as the state monitoring table, The force sensor output value range is wide (because it changes from the non-contact state to the contact state) as in the fitting / insertion work, and the sensor output value pattern changes within the above range due to the position error of the robot or parts. In some cases, it is not applicable.

  Since Patent Document 2 holds a robot position (axis parameter) and a force sensor value as an operation correction table, it is determined whether or not the work is good due to a change in the absolute value of the force sensor according to the position of the robot. Although it becomes possible, there is a problem that the quality of work due to a cause that does not affect the force sensor absolute value cannot be determined. In addition, when the motion correction table is automatically created at the teaching stage, the robot is manually operated to manually create work conditions (good / bad) such as interference points and interference avoidance points based on human judgment. It does not reflect the state when the robot is in operation. Therefore, there is also a problem that the quality of work cannot be determined accurately.

In order to solve the above problems, the present invention has structural features as described below.
Of course, the description of the scope of claims can be amended based on the description of the specification at the time of filing the application, but at that time, the contents of the following “inventions” are not scheduled to be corrected. (Each "invention" below has the meaning as a gist of disclosure in the present specification).

According to the first aspect of the present invention, in the step of assembling the gripped part by the hand provided on the robot arm of the industrial robot, a plurality of fitting / inserting operations associated with the operation ID are repeatedly performed in advance. Is a system for determining the work result of the actual work associated with the work ID and estimating the cause of failure by organizing the work quality and cause of failure stored in the force sensor for detecting the force acting on the hand Each physical quantity corresponding to one or more feature quantity parameters defined in advance based on a force sensor signal outputted from the robot and / or a position detection signal outputted from each joint axis drive motor provided in the robot arm. Feature quantity extraction means (113) that extracts as a quantity, and for each work ID, a work that indicates each extracted feature quantity and either work success or work failure A history data recording means (114; 116) for recording first history data including pass / fail and cause of work failure, and a feature quantity existing area recording means (123), the history data recording means (114; 116). Based on the first history data stored and associated with each work ID, an optimum feature parameter set is selected from one or more feature parameters and each selected feature parameter is represented as a coordinate axis. And a feature quantity existence area recording means (123) for determining and recording a feature quantity existence area in the coordinate system based on the feature quantity existence range of the selected optimum feature quantity parameter. Each feature quantity existence area is associated with a work pass or fail cause indicating work success included in the first history data, and an optimal feature parameter set. DOO is characterized by being selected on the basis of the overlapping regions of the feature quantity present area associated with the work quality showing a work success, and the feature amounts exist regions associated with the cause of work fails, thereby, The work pass / fail judgment means (118) judges the work result of the actual work associated with the work ID and estimates the cause of failure based on the feature quantity existence area recorded in the feature quantity existence area recording means (123). Make it possible.

  In the invention according to claim 2, the feature amount existence region based on the feature amount existence range of the feature amount parameter is an average of the feature amounts of the first history data recorded in the history record data means (114; 116). Based on the value and the variance, it is calculated to have a pre-specified existence probability.

In the invention according to claim 3, the work pass / fail judgment means (118) further selects the optimum work selected from the feature quantities extracted by the feature quantity extraction means (113) for the actual work associated with the work ID. When the plot in the coordinate system of the feature quantity for the feature quantity parameter is included only in the feature quantity existence area associated with the work quality indicating the work success, it is determined that the work is successful, and is associated with the cause of the work failure. if it included only in one or more of the feature quantity present in the region, as well as determined that the work failed to extract the failure cause of the feature amount existence region are associated, or associated with the work quality showing a work success Included in one or more of the feature quantity existence areas associated with the specified feature quantity existence area and the cause of the work failure, or associated with work quality indicating work success If it is not included in any of the feature quantity existence areas and the feature quantity existence areas associated with the cause of the work failure, it is determined that the judgment is impossible, and is displayed by the work quality / cause display means (119) provided in the system. Is done.

In the invention according to claim 4, the work quality determination means (118) further determines the plot when the plot is included only in two or more of the feature amount existence regions associated with the cause of the work failure. comprise feature quantity existing region has is to extract all of the two or more failure cause is associated, two or more failure cause is displayed by the work quality, cause the display means (119).

In the invention according to claim 5, in a case where the work quality determination unit (118) is determined to work successfully, a feature amount present area associated with the work quality showing a work success, associated with the cause of the working failure and if there are duplicate region and the characteristic values existing region, based on the size of the overlapping area size and the task feature quantity present area associated with the work quality indicating success, certainly in the determination of when working success The likelihood is calculated, and the work quality / cause display means (119) further displays the probability of determination.

  In the invention according to claim 6, the one or more feature amount parameters for the feature amount extracted by the plurality of feature amount extracting means (113) are the insertion amount of the gripping component calculated from the position detection signal. , Peak value of time differentiation in force sensor signal, work amount at insertion calculated from insertion amount and force sensor signal, and change in posture angle of hand calculated from position detection signal of each joint axis drive motor of robot arm Including any one or more of the quantities.

In the invention according to claim 7, the history record data means (114; 116) further includes the feature quantity extracted by the feature quantity extraction means (113) and the work quality determination means for the actual work associated with the work ID. The second history data including the work quality determined by (118) and the estimated cause of failure is additionally recorded, and the feature amount existence area recording unit (123) includes a history data recording unit (123). 114; 116), each feature amount existence region can be updated based on the first history data and the second history data.

In the invention according to claim 8, the optimum feature parameter set is further associated with the quality of the work indicating the work success from the predetermined number of feature parameter parameters among the one or more feature parameter. A set of feature amount parameters is selected in which there is no overlapping region between the feature amount existing region and each feature amount existing region associated with the cause of the work failure or the overlapping region is the smallest.

In the invention according to claim 9, further, when the set of the feature parameters overlap region is not present there is more than one associated with the cause of the work fails characterized abundance area associated with the work quality showing a work success An optimum feature parameter set is selected based on the distance between the centers of the feature parameter existing regions.

In the invention according to claim 10, when there is no feature parameter set that does not have an overlapping region, the predetermined number is incremented, and the successful operation is performed from the incremented feature parameter set. A feature parameter set is selected in which there is no overlap area between the feature quantity existence area associated with the indicated work quality and each feature quantity existence area associated with the cause of the work failure or the overlap area is the smallest.

  According to an embodiment of the present invention, in the test operation process, a plurality of data including history data relating to one or more feature amount parameters and work quality and failure cause history data obtained by judgment of the teaching coordinator during the test operation. Are stored in the table, and at that time, the feature amount existence range at the time of work success and the feature amount existence range at the time of work failure are calculated for each cause of failure.

  Thereby, in the actual operation process, based on the calculated feature amount existence range at the time of successful work and the feature amount existence range at the time of work failure, it is possible to accurately determine the quality of work depending on various causes. It is what you play.

Further, according to the present invention, when the operation fails, the robot is stopped, and the semi-finished product having the insertion failure can be prevented from flowing to the subsequent process. Since the cause of the failure is also presented to the line supervisor, the production line can be quickly restored.
Furthermore, the present invention calculates and displays the probability of determination at the time of successful work in accordance with the degree of overlap of the feature amount existence ranges, so that the determination reliability can be objectively grasped.

Furthermore, the present invention also has an effect that the quality of work depending on various causes can be accurately determined by selecting the optimum feature parameter from the history data during the test operation.
Furthermore, the present invention also has an effect of accurately determining work quality depending on various causes by feeding back data including work quality and failure cause acquired at the time of work as history data.

Configuration diagram of a general working robot in the prior art Configuration diagram of impedance controlled industrial robot in the prior art Block diagram of impedance control in the prior art The figure explaining the quality (success / failure) of fitting / inserting work in impedance control of the prior art The block diagram of the work quality determination apparatus at the time of the fitting / insertion work of the Example of this invention The figure explaining the time response of the amount of movement (insertion) at the time of fitting and insertion work of the example of the present invention, and acting force The flowchart explaining the whole step which judges the quality of work in the process of assembling the gripping parts of the example of the present invention, and presumes the cause of failure. The figure explaining the feature-value log | history table and work quality / cause log | history table of an Example of this invention The figure explaining the tendency of the work failure situation 1 of the Example of this invention, and its feature-value The figure explaining the tendency of the work failure situation 2 of the Example of this invention, and its feature-value The figure explaining the tendency of the work failure situation 3 of the Example of this invention, and its feature-value The figure explaining the tendency of the work failure situation 4 of the Example of this invention, and its feature-value The figure explaining the tendency of the number (type) of feature-values of the Example of this invention and the probability of work quality determination The flowchart explaining the step which selects the optimal feature-value parameter in the Example of this invention. The figure explaining the feature-value existence range at the time of work success in the Example of this invention, and the feature-value existence range at the time of work failure

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  In this embodiment, as assembly of a gripping work by a robot hand, it is possible to determine work quality (success / failure) at the time of fitting / insertion work, and to estimate the cause of work failure in the case of work failure This is a robot system that performs work result determination.

FIG. 5 is a configuration diagram of a work quality determination apparatus during fitting / insertion work according to the present invention. In this figure, the same reference numerals are assigned to the same names as those in FIGS.
In FIG. 5, the work quality determination device 112 is configured to be implemented in the robot controller 102, that is, a feature amount extraction unit 113, a feature amount history recording unit 114, a work quality / cause history recording unit 116, and a feature amount existence range calculation unit 117. , And a part related to the work quality determination unit 118 and a configuration mounted on the portable teaching operation panel 103, that is, a part related to the work quality / cause input unit 115 and the work quality / cause display unit 119.

  In the present embodiment, each of the above components is included in the same work quality determination device 112, but is not limited to this, for example, each component is included in a separate device, It can also be configured to pass data between devices.

The work quality determination device 112 can cause each of the components 113 to 119 to execute processing described below by a processor included in the device 112.
The feature amount extraction unit 113 extracts a feature amount based on the feature amount parameter from the force sensor signal and the position detection signal of each joint axis drive motor of the robot arm.

  In the work result determination system according to the present embodiment, there are one or more feature amount parameters, and an arbitrary number of parameter sets can be selected in advance. Specific feature parameter and its extraction method will be described later.

  The feature amount history recording unit 114 stores feature amount history data when the operation is repeatedly performed repeatedly during the test operation. The work quality / cause input unit 115 receives an input of work quality or cause of failure determined by a human (teaching coordinator) when the work is repeatedly executed during the test operation. Information can be input by operating the keys 103a and the display screen 103b (touch panel type) of the portable teaching operation panel 103. The work quality / cause history recording unit 116 records the work quality / failure cause history data input from the work quality / cause input unit 115 together with the program execution step ID and the work date and time. The feature amount existence range calculation unit 117 uses the history data recorded in the feature amount history recording unit 114 and the history data recorded in the work quality / cause history recording unit 116 to determine the feature amount at the time of work success and work failure. The feature value existence range based on the parameter is calculated. The work pass / fail determination unit 118 is based on the feature amount obtained from the feature amount extraction unit 113 at the time of performing the work and a plurality of existence ranges calculated by the feature amount existence range calculation unit 117 and recorded in the feature amount existence range recording unit 122. The feature amount existence area is determined and recorded in the feature amount existence area recording unit 123, and the success or failure of the work and the cause thereof are determined. A detailed method for determining whether the work is good or bad will be described later. The work quality / cause display unit 119 can display the work quality / failure cause determined by the work quality determination unit 118 on the display screen 103 b of the portable teaching operation panel 103. Based on the display, the production line supervisor can visually grasp the quality of the work and can quickly restore the line based on the displayed cause of failure when the work fails.

Next, the operation principle of the work quality determination device described above will be described in detail.
FIG. 6 shows the time response of the acting force Ffb in the insertion direction (Z direction) and the movement amount Pfb when the robot 101 is in the impedance control state and the operation program execution unit 109 executes the fitting / insertion operation. At time T1, the fitting part 110a and the fitting part 110b come into contact with each other, and the insertion starts after time T1, and at time T2, the fitting part 110a or the fitting part 110b comes into contact with the bottom of the hole. Insertion is completed at time T3. In the feature amount extraction unit 113 in FIG. 5, from the response of the position (movement amount) and force (acting force) shown in FIG. 6, the insertion amount L, the peak value of the time derivative of the acting force, the work amount W at the time of insertion, A feature quantity based on a feature quantity parameter such as a change amount of a posture angle is extracted. The insertion amount L is calculated by subtracting the movement amount Z1 at time T1 from the movement amount Z2 at time T3. For the peak value of the time derivative of the acting force, the time derivative of the acting force is obtained and the maximum value is calculated. Regarding the work amount W at the time of insertion, first, the amount of movement is differentiated with respect to time to obtain the insertion speed and is calculated based on the following equation.

Work W = ∫ (insertion speed × acting force) dt (3)
In addition, the integration interval of Formula (3) is from time T1 to T3. The change amount of the posture angle is calculated by subtracting the posture angle at the start of insertion (time T1) from the posture angle of the gripping work (fitting part 110a) at the time of completion of insertion (time T3).

In the embodiments of the present invention, the insertion amount, the force differential peak, the work amount, and the posture angle are described as the feature amount parameters, but the present invention is not limited to this.
Next, with reference to FIG. 7, an overall process for determining the quality of work and estimating the cause of failure in the process of assembling the gripping part according to the embodiment of the present invention will be described in the order of steps in time series.

  The overall process is largely composed of a test operation process (S701 to S703) and an actual work process (S706 and S707). Further, when starting the actual work process after the test operation process, it is necessary to calculate the existence range of the feature quantity (S704) and define the feature quantity existence area (S705). This process will be described later. It is also possible to perform it dynamically in the actual work process.

  In the test operation process, first, a physical quantity corresponding to one or more feature quantity parameters defined in advance based on each detection signal is extracted as each feature quantity by the feature quantity extraction unit 113 (S701). Each of the extracted feature values is recorded in the feature value history recording unit 114 (S702), and the work quality indicating either work success or work failure and the cause of the work failure are input from the work quality / cause input unit. Is recorded in the work quality / cause history recording unit 116 (S703). During the test operation process, these steps (S701 to S703) are repeatedly executed.

  After the test driving process ends, the feature quantity existence range calculation unit 117 calculates the existence range of each feature quantity based on the history data recorded in the test driving process, and records it in the feature quantity existence range recording unit 122 ( Then, a feature amount existence region for the feature amount parameter set is defined and determined, and recorded in the feature amount existence region recording unit 123 (S705).

  Thereafter, in the actual work process, the work quality determination unit 118 determines the work result of each actual work and estimates the cause of failure based on the feature amount existence area (S706). This work determination result is displayed on the work quality / cause display section 119 (S707).

  In the feature quantity history recording unit 114 of FIG. 5, as shown in FIG. 8A, each feature quantity at the time of work execution obtained from the feature quantity extraction unit 113 is converted into an execution step ID, date and time (date / date) of the operation program. Time). In the work quality / cause input unit 115, when the teaching coordinator of the robot system repeatedly performs the fitting / insertion work as a test operation, the result of the teaching coordinator determining the quality of the work and the cause of failure in case of failure Is input from the portable teaching operation panel. As shown in FIG. 8B, the work quality / cause history recording unit 116 indicates the work quality / failure cause input from the work quality / cause input unit 115, the execution step ID, date / time (date / time) of the operation program. ). The feature quantity corresponding to each row of the table data in FIG. 8B can be obtained by searching the row of the same execution step ID and the same date and time from the table data in FIG. For convenience of explanation, the table data is divided into (a) and (b). However, they can be integrated and recorded as table data of one history data recording means.

  Note that the program execution step ID (“00100” in the case of FIG. 8B) is numbered in integer units that increase by 1 for each instruction (MOVE command, etc.), with the top NOP of the operation program being 00000. It is. Although the specific contents of the operation program are not shown, “NOP” means “No Operation”, and is an instruction that is always given to indicate the beginning of the operation program. In addition to the MOVE command for operating the robot, the operation program after the NOP is configured by an input / output command to the external device connected to the robot controller 102, a conditional branch command, and the like.

  When the repetitive mating / insertion work is executed as a test operation during the test operation process, the same program execution step ID is assigned. Each work in the actual work process is associated with the program execution step ID. Further, each table / data of the fitting / inserting work stored in the work quality / cause history recording unit 116 can be uniquely specified by the program execution step ID and the date / time (date / time). A work ID that identifies the work is configured.

  In a normal case, the test operation is performed several tens to several hundreds per program execution step ID. If there is a defect, adjust it each time so that the success rate is almost 100% (99.9% or more).

Next, details of the feature amount existence range calculation unit 117 will be described.
First, for the fitting / insertion work shown in FIGS. 4C to 4E, how the feature amount existence range at the time of the work success and the feature amount existence range at the time of the operation failure are calculated, and these feature amount existences are calculated. Whether a region is defined will be described with reference to specific examples shown in FIGS.

  In the work result determination system of the present embodiment, the definition of the feature amount existence region is executed by the processor included in the work quality determination device 112 when the actual operation is started after the test operation is completed.

It is assumed that a feature parameter set to be adopted is set in advance when defining the feature parameter existence region.
FIG. 9A shows an example of a situation where the fitting / inserting operation fails (cause of failure 1). This is an example in which fitting / insertion has failed because the air pressure of the workpiece fixing jig 111 has dropped, and the position of the fitted part 110b has slightly shifted from the time of teaching. FIG. 9B shows a plot of feature value values of the failure example together with a plot at the time of success. Here, as an example of the feature amount parameter, an insertion amount (feature amount 1) and a peak value of time derivative of the force (feature amount 2) are adopted. In FIG. 9B, a plot 501 is a plot of feature value at the time of success, and a plot 502a is a plot of feature value at the time of failure. The reason why the feature value plot 502a at the time of failure has a larger peak value of the time differential of the force than the plot 501 at the time of success is that the position of the workpiece 110b is slightly shifted and the impact at the time of contact is increased. is there.

  In this case, the feature quantity existence range calculation unit 117 extracts from the history table data shown in FIG. 8B what is determined to be “successful” for the same work (same execution step ID “00100”). The feature value of the corresponding feature parameter (feature parameter 1 and feature parameter 2) is acquired by referring to the history table data 8 (a), and the average value and standard deviation (variance) value for each feature parameter Is calculated as the feature amount existence range. Then, from the average value and the variance value, the inside of the rectangular area 503 in FIG. 9B is defined as the feature amount existence area 503 at the time of success.

  Here, the “feature amount existence range” means, for example, a feature value range (from / to) corresponding to the feature parameter for the feature parameter, but is not limited thereto. In the above example, the feature amount existence area is defined as a rectangular region based on the feature amount existence range. However, the feature amount existence region is not limited to this, and is defined based on an ellipse or the like in a two-dimensional coordinate system. May be.

  In FIG. 9B, specifically, for the horizontal axis (feature value 1) direction of the feature quantity existence region 503, “average value of feature quantity 1 ± α × variance value of feature quantity 1”, vertical axis (feature For the amount 2) direction, the existence area (upper limit and lower limit) is calculated as “average value of feature amount 2 ± α × variance value of feature amount 2”. Here, α is a coefficient for setting the existence probability, and the existence range of the feature amount is calculated so as to have a pre-specified existence probability. For example, assuming that the plot has a normal distribution, if α = 2, the existence probability is 95.44%, and if α = 3, the existence probability is 99.73%.

  For the work failure, the history table data in FIG. 8B lists the same work (same execution step ID “00100”) determined as “failure” for the same cause, and FIG. Refer to the history table data in a) to obtain all the feature value values of the corresponding feature parameter (feature value 1 and feature value 2), and calculate the average value and standard deviation (variance) value for each feature value. To do. From the average value and the variance value, the feature amount existence range at the time of failure is calculated in the same procedure, and the inside of the rectangular area indicated by 504a in FIG. 9B is defined as the feature amount existence region 504a at the time of failure. Regarding the feature quantity existence range at the time of failure, the feature quantity existence range is obtained for each cause of failure. Each calculated existence range related to the work success and the work failure is recorded in a feature quantity existence range recording unit 122 (not shown) included in the feature quantity existence range calculation unit 117 as table data shown in FIG. 8C. .

  Note that the “feature amount existence range” portion of each table data illustrated in FIG. 8C in the feature amount existence range recording unit 122 includes, for example, the feature amount parameter and the feature amount value corresponding to the feature amount parameter. A set of ranges (from / to) can be stored, but is not limited to this. Further, the feature amount existence area defined based on the feature amount existence range can be further stored in the feature amount existence area recording unit 123.

  The work quality determination unit 118 in FIG. 5 includes a feature amount plot obtained from the feature amount extraction unit 113 during actual work execution and a feature amount existence region based on the feature amount existence range recorded in FIG. The comparison is made and it is checked in which feature quantity existing region. Then, when the obtained feature amount plot is only in the feature amount existence region at the time of success, it is determined as “success”, and when it is only in the feature amount existence region at the time of failure, it is determined as “failure”. If it belongs to both of the feature quantity existence regions or does not belong to either of them, it is determined that “determination is impossible”. If it is determined as “failure”, the cause of failure is searched from the table of FIG. Further, when the feature amount plot is included in any of the feature amount existence areas for two or more failure causes, the two or more failure causes are searched.

  The work quality / cause display unit 119 in FIG. 5 displays the work quality (success / failure) and cause (only when failure) obtained by the work quality judgment unit 118 on the display screen 103 b of the portable teaching operation panel 103. Further, when the work quality determination unit 118 determines that the work has failed, the robot 101 is stopped by operating the operation program execution unit 109, so that it is possible to prevent a semi-finished product that has been poorly inserted from flowing into a subsequent process. Furthermore, since the supervisor of the production line on which the robot 101 is operating can grasp the cause of the stop of the robot by looking at the work quality / cause display unit 119, the production line can be quickly restored.

  10 to 12 show examples of failure situations (failure causes 2 to 4) different from those in FIG. 9 and plots of their feature values. The failure cause and the tendency of the feature amount will be described in the order of FIGS.

  FIG. 10A shows a situation where the latch 110c is not fully fitted when the latched portion 110b has the latch 110c. Here, an insertion amount (feature amount 1) and a peak value of time differentiation of the force (feature amount 2) are employed as the feature amount parameters (FIG. 10B). When the latch is insufficient, the time change of the force becomes slower than when the latch is sufficient. Therefore, as shown in FIG. 10B, the plot 502b of the feature value at the time of failure is a plot 501 at the time of success. It shows a tendency that the peak value of time derivative of force is smaller than that. The feature amount existence region 504b is a feature amount existence region at the time of failure (failure cause 2) defined based on the feature amount existence range calculated by the feature amount existence range calculation unit 117. The difference in insertion amount between when the latch is sufficient and when the latch is insufficient is very small, and when the parts and hand claws are flexible (for example, made of plastic), the variation in the insertion amount increases. Disappear.

  FIG. 11A shows that when the fitted component 110b is set on the workpiece fixing jig 111 in a tilted state, the fitted component 110a is not inserted only by rubbing the side surface of the fitted component 110b. Indicates the situation. Here, the insertion amount (feature amount 1) and the amount of work at the time of insertion (insertion amount 3) are adopted as the feature amount parameters (FIG. 11B). When the part rubs the side surface, the acting force (resistance force) is smaller than when it is normally inserted. Therefore, as shown in FIG. The work amount at the time of insertion tends to be smaller than the plot 501 at the time of success. The feature amount existence area 504c is a feature amount existence range at the time of failure (failure cause 3) calculated by the feature amount existence range calculation unit 117 and defined based on this. Even when the side is not rubbed only by rubbing, the insertion amount (movement amount) calculated from the position of the hand 104 as shown in FIG. 11A cannot be distinguished from the case where it is correctly inserted. There is.

  FIG. 12 (a) shows a situation where a flexible case as a fitting part 110a is not fully fitted because a foreign object is caught in one of the guides 110d when the flexible case is fitted and inserted into a fitting part 110b with a guide 110d. ing. Here, the insertion amount (feature amount 1) and the change amount of the hand posture angle are adopted as the feature amount parameters (FIG. 12B). When a foreign object is caught in one of the guides 110d, it tilts slightly obliquely. Therefore, as shown in FIG. Shows a tendency to increase. The feature amount existence region 504d is a feature amount existence region at the time of failure (failure cause 4) calculated by the feature amount existence range calculation unit 117 and defined based on this. In the case of the flexible case, the case is deflected by applying a force for insertion, so that it is difficult to distinguish from the success case only by the insertion amount (calculated from the position of the hand 104) as in the example of FIG.

  Each of the failure situations shown in FIGS. 9 to 12 is an example in which it is impossible to accurately distinguish between success and failure only with one feature parameter called insertion amount. By increasing the number of feature parameters such as the amount of work and the amount of change in the posture angle of the hand, various failure causes can be accurately distinguished. In addition, one feature parameter is not necessarily required for one cause of failure, and as the number of feature space dimensions (number of features) N increases, the number of quadrants increases by 2 to the Nth power. It is possible to distinguish the cause of failure up to 2 N power.

  The example shown in FIGS. 9 to 12 shows a situation where the feature amount existence area 503 at the time of success and the feature amount existence area 504a at the time of failure can be reliably distinguished from each other by two feature amounts. Has an overlapping area in the existence area at the time of success and the existence area at the time of failure, and there is a case where clear distinction is difficult (FIG. 13B). Even if there are overlapping portions of such areas, the work quality determination unit 118 determines “success” when the feature value obtained at the time of performing the work is only within the feature value existing range at the time of success.

  At this time, as shown in FIG. 13, the work pass / fail determination unit 118 determines whether or not the determination is based on the overlapping degree of the areas based on the size of the overlapping area and the size of the feature amount existing area when the work is successful. It is also possible to calculate “property” (whether clear distinction can be made) and to display the likelihood on the work quality / cause display unit 119. In FIG. 13, a feature amount existence area 503 is a feature amount existence range at the time of work success calculated by the feature amount existence range calculation unit 117. Similarly, a feature amount existence area 504 is a feature amount existence area and a feature at the time of work failure. The quantity plot 505 is a plot of feature values at the time of work execution. FIGS. 13A to 13C are examples of 1 to 3D feature spaces, respectively. In the case of FIG. 13A, since 30% of the feature amount existence region at the time of success overlaps with the feature amount existence region at the time of failure, the feature amount plot 505 is determined as “successful”. "Is calculated as 70%. Similarly, the “probability” of the success determination in the case of FIG. 13B is calculated as 90%, and the “probability” of the success determination in FIG. 13C is calculated as 100%. FIG. 13 also shows that by increasing the number of feature parameters, there is no overlap between the feature amount existing area at the time of work success and the existing range area at the time of work failure, and the work quality can be clearly distinguished. Yes.

  According to the present embodiment, by increasing the number of feature parameter used to determine whether the work is good or not, it is possible to minimize the overlap between the feature amount existence area at the time of work success and the existence range area at the time of work failure. (Success / failure) can be clearly distinguished, and the accuracy of the estimated cause of failure can be improved.

  In the robot system according to the present embodiment, the quality of work at the time of work execution is determined based on the history data recorded in the feature quantity history recording unit 114 and the work quality / cause history recording unit 116 for a predetermined number of feature quantity parameters. It is possible to select a feature parameter set that is an optimal feature parameter.

  The work quality determination device 112 in the work result determination system of this embodiment is coupled to the feature amount history recording unit 114, the work quality / cause history recording unit 116, the feature amount existence range calculation unit 117, and the feature amount existence region recording unit 123. And a feature parameter selection unit 120 (not shown). The feature parameter selection unit 120 (not shown) includes the history data recorded in the feature value history recording unit 114 and the work quality / cause history recording unit 116, and the feature parameter calculated by the feature value existence range calculation unit 117. Based on the existence range of the feature quantity based on, a set of feature quantity parameters to be an optimum feature quantity parameter in the work quality determination at the time of performing the work is selected. In addition, a set of optimum feature parameters selected by the feature parameter selection unit 120 defines a coordinate system that employs the selected optimum feature parameters as coordinate axes, and a feature value existing area in the coordinate system Can be determined and recorded in the feature amount existence area recording unit 123.

  Here, the feature amount existence areas in the coordinate system are associated with work quality at the time of work success or work failure, respectively. Then, the optimum feature parameter set is selected based on the overlap region between the feature value existing region at the time of successful work and each feature value existing region associated with the cause of failure at the time of work failure.

The system also includes a feature parameter recording unit 121 (not shown) for storing process data in the optimum feature parameter determination process.
The optimum feature parameter determination processing is executed by a processor included in the work quality determination device 112 when starting actual work after the test operation is completed. During actual work, the work quality judgment device 118 judges work quality based on the feature quantity parameter set selected by the optimum feature quantity parameter selection process.

  The optimum feature parameter selection process will be described in detail below with reference to FIG. It is assumed that the number of feature amount parameters (that is, the number of dimensions) is set in advance. In the following description, the number of feature parameter parameters is “2”, and a two-dimensional coordinate system in which two feature parameter sets are used as coordinate axes is defined and described. However, the present invention is not limited to this.

  First, two feature parameter sets are arbitrarily selected (S1401). In the example of FIG. 8A, the feature amount history recording unit 114 has four feature amount parameters (feature amounts 1 to 4) of an insertion amount, a force differential peak, a work amount, and a posture angle. One set will be selected. Next, based on the history data recorded in the feature amount history recording unit 114 and the work quality / cause history recording unit 116 using the selected set of feature amount parameters, the feature amount existence range calculation unit 117 performs “success”. The existence range of “” and the existence range of “failure” are respectively calculated to define each existence region (S1402 and S1403). When there are a plurality of failure causes for the existence range of “failure”, the existence range is calculated for each failure cause. Next, it is determined whether or not there is an overlapping area in the existing areas of “success” and “failure” (S1404). If there is an overlapping area, the “probability” described above is calculated and stored in the feature parameter recording unit 121 together with the set of selected feature parameters (S1405). If there is no overlapping area, the “probability” is set to 100% and stored together with the feature parameter set (S1406). These steps from S1402 to S1404 are performed for all feature parameter sets (S1407). Next, on the basis of all six “probability” values stored in the feature parameter recording unit 121, an optimum feature parameter set is selected, which is optimal for the feature existing region recording unit (123). A feature amount existence area corresponding to the set of feature amount parameters is recorded (S1408). As the set of optimal feature amount parameters, for example, the one with the maximum “probability” value can be selected.

  FIG. 15A shows a “success” work area 1503 and a “failure” work area 1504 when feature quantity 1 and feature quantity 2 are selected as the feature quantity parameters. Similarly, FIG. 15B shows a case where feature amount 3 and feature amount 4 are selected. In the case of FIG. 15A, the “probability” value is stored as 90% in the feature parameter recording unit 121 (S1405). On the other hand, in the case of FIG. 15B, the “probability” value is stored as 100% in the feature parameter recording unit 121 (S1406). Therefore, as an optimal feature parameter set, the feature regions exist in areas 1503 and 1504 do not overlap (the value of “probability” is 100%), and the feature as shown in FIG. The quantities 3 and 4 are selected, and the feature quantity existence areas 1503 and 1504 corresponding to the set of the feature quantities 3 and 4 are recorded in the feature quantity existence area recording unit (123) (S1408).

  Note that when there are a plurality of feature parameter sets whose “probability” value is 100%, for example, the center of the “success” work area 1503 and “failure” as shown in FIG. The distance d between the center of the work area 1504 and the center of the work area 1504 may be calculated, and the feature parameter set with the larger value of the distance d may be selected as the optimum feature parameter parameter set. In addition, as shown in FIG. 15D, the size of the “success” work area 503 (for example, area f1 × f2) and the size of the “failure” work area 504 (for example, area g1 × g2) are set. Based on this, for example, a feature parameter set having the smallest area ratio (g1 × g2) / (f1 × f2) may be selected as the optimum feature parameter set.

Further, FIG. 15 illustrates the case of one failure cause, but the same applies to cases where there are a plurality of failure causes.
Furthermore, in this embodiment, the number of feature parameter (ie, the number of dimensions) is set in advance, but in another embodiment, there is no overlapping area while the number of dimensions is incremented. You may make it select the optimal feature-value parameter set. For example, assuming that the initial number of dimensions is 1, and there is no feature parameter set that does not have an overlapping region when selecting an optimal feature parameter set in S1408 (that is, a feature whose “probability” value is 100%. If the quantity parameter set is not detected), the number of dimensions is incremented to 2 and the feature quantity parameter set in which there is no overlapping area (that is, the feature quantity parameter with a “probability” value of 100%) Search). As a result, it is not always necessary to use the maximum number of feature parameter sets in actual work, and a minimum feature parameter set can be used to perform work pass / fail judgment and estimation, improving processing efficiency. Can be made.

  As described above, the data of the feature amount existing area related to the selected optimum feature amount parameter is recorded in the feature amount existing area recording unit 123 as the table data together with the program execution step ID and the work quality.

  According to the present embodiment, the optimum feature amount in the work quality determination at the time of performing the work based on the history data recorded in the feature amount history recording unit 114 and the work quality / cause history recording unit 116 with respect to the number of feature amount parameters. It is possible to select a feature parameter set as a parameter. As a result, the feature amount existence region becomes more accurate, and accordingly, the quality of the work determined in actual work and the accuracy of the estimated cause of failure can be increased.

  In the robot system of the present embodiment, the data related to the work quality determined by the work quality determination unit 118 at the time of performing the work can be additionally recorded sequentially in the feature amount history recording unit 114 and the work quality / cause history recording unit 116. The feature amount extracted by the feature amount extraction means (113) at the time of performing the work, the work quality determined by the work quality determination means (118), and data including the estimated cause of failure can be fed back as history data. . That is, the feature amount history recording unit 114 records not only the test operation process but also the feature amount extracted by the feature amount extraction means (113) at the time of performing the work, and the work quality / cause history recording unit 116. Stores not only the quality of work and the cause of failure input and stored from the work quality / cause input unit 115 during the test operation, but also the quality of work determined by feedback from the quality of work determination unit 118 and the estimated cause of failure. To do. At this time, the same execution step number (in this case, “00100”) is stored in the “program execution step ID” of FIG. 8B, and similarly, the date on which the work is executed is stored in the “date / time”. In addition, the quality of work and the cause of failure determined at the time of work execution are stored in the time, “work quality” and “cause of failure”.

  Then, the feature quantity existence range calculation unit 117 sequentially or periodically calculates the feature quantity existence range at the time of “success” work and the feature quantity existence range at the time of “failure” work, and periodically defines each feature quantity region. The data stored in the feature quantity existence range recording unit 122 and the feature quantity existence region recording unit 123 can be updated again.

  Note that the feature amount existence range calculation unit 117 calculates the feature amount existence range calculation timing (sequentially or regularly) through the portable teaching operation panel 103 or is set during actual work, and determines whether the work is good or bad. Controlled by the processor of device 112.

According to the present embodiment, the feature amount existence region becomes more accurate, and accordingly, the accuracy of the determined work quality and the estimated cause of failure can be increased.
As described above, according to the present invention, the history data relating to a plurality of feature quantities, and the feature quantity existence range at the time of successful work from the work pass / fail and failure cause history data obtained by the judgment coordinator during the test operation, Since the feature amount existence range at the time of work failure is calculated for each cause of failure, the work quality depending on various causes can be accurately determined.

In addition, the present invention presents the cause of the failure to the line supervisor at the time of work failure, so that the production line can be quickly recovered.
Furthermore, since the present invention calculates and displays the determination probability according to the overlapping degree of the feature amount existence ranges, it is possible to objectively grasp the determination reliability.

  Furthermore, the present invention selects the optimum feature parameter set from the history data during the test operation, so that it is possible to accurately determine whether the work depends on various causes without depending on the entire feature parameter set. There is an effect that it can be determined.

  Furthermore, the present invention has an effect that it is possible to accurately determine the quality of work depending on various causes by feeding back data including the quality of work and the cause of failure acquired at the time of work as history data.

  With the work quality determination device at the time of the fitting / insertion work of the present invention, various assembly line processes such as automobile parts and home appliances can be automated using a robot without reducing productivity.

  In the present specification, the “robot” refers to a device that performs some work on behalf of a person. A typical example is an industrial robot that performs automatic work continuously autonomously to some extent, but is not limited thereto. “Robots (devices that perform some work on behalf of humans)” to be developed in the future, including all “robots” that can use the invention included in the technical scope of the claims of the present application And

  Particular embodiments described herein have been described. Other embodiments are within the scope of the claims. For example, the actions recited in the claims can be performed in a different order and still achieve desirable results. By way of example, the processes shown in the accompanying drawings do not necessarily require the particular order shown or sequential order to achieve desirable results.

  Moreover, the function which one component has may be implement | achieved by two or more physical structures, and the function which two or more component elements have may be implement | achieved by one physical structure. The invention of the system can be grasped as an invention of a method in which the functions of each component are sequentially executed, and vice versa. In the method invention, the steps are not limited to being performed in the order described, but can be performed in any order as long as the overall function can be performed consistently. . These inventions are also established as a program that realizes a predetermined function in cooperation with predetermined hardware, and also as a recording medium that records the program. The present invention can also be realized as a computer data signal embodied on a carrier wave and having a program code.

Other embodiments of the present aspect include corresponding systems, apparatuses, devices, computer program products, and computer readable media.
The present invention can be corrected to the various aspects described above by amendment of procedures during examination and within the scope of legal restrictions in a trial for correction or a request for correction after patent.

  In addition, the judgment of “substantially change the scope of claims” in a trial for correction after a patent or a request for correction is based on whether or not a new component has been added to the claims at the time of patent (that is, because of so-called external addition). Or whether it further restricts any one or more components of the claims at the time of the patent (ie, so-called internal additions have been made) Therefore, it should be made from the viewpoint of whether the effects of the claimed invention before and after the correction are similar.

  This specification includes a number of specifics, which should not be construed as a limitation on the scope of what is claimed or can be claimed, but a description of features specific to a particular embodiment. Should be interpreted as

In the context of separate embodiments, certain features described herein can be further implemented in combination in a single embodiment.
In contrast, the various features described in the context of a single embodiment can be further implemented in a number of embodiments or in any suitable subcombination.

  Further, features are described above to act in a particular combination and may be initially so claimed, but one or more features from the claimed combination may be several In that case, it can be carried out from that combination, and the claimed combination can be directed to a small coupling or various small couplings.

  Similarly, operations are illustrated in a particular order, which means that such operations are performed in the particular order shown or in sequential order to achieve the desired results, or their It should not be understood as requiring that all illustrated operations be performed.

DESCRIPTION OF SYMBOLS 101 Robot 102 Controller 103 Portable teaching operation panel 103a Operation button group 103b Display screen 104 End effector 105 Force sensor 106 Impedance control unit 106a Position control system 106b Impedance control calculation unit 106c Impedance model 106d Speed resolution calculation unit 107 Actuator drive amplifier unit 108 Operation program storage unit 109 Operation program execution unit 110a Fitting component 110b Fitted component 110c Latch 110d Guide 111 Work fixing jig 112 Work quality determination device 113 Feature quantity extraction unit 114 Feature quantity history recording unit 115 Work quality / cause input unit 116 Work quality / cause history recording unit 117 Feature quantity existence range calculation unit 118 Work quality judgment unit 119 Work quality / cause display unit 123 Feature quantity existence region recording unit 301 Time response 302 of acting force at the time of performing the entry work Time response 501 of the amount of movement (insertion) at the time of execution of the fitting / insertion work Plot 502a of the feature value at the time of successful work 502a of the feature value at the work failure situation 1 Plot 502b Plot of feature value in work failure situation 2 502c Plot of feature value in work failure situation 3 502d Plot of feature value in work failure situation 4 503 Existence range of feature quantity at work success 504 Feature feature existence range 504a Work failure status 1 Feature value existence range 504b Feature failure presence range 504c Feature failure existence range 504c Feature failure existence range 504d Work failure status 504d Work failure status Feature quantity existence range 505 in 4 Plot of feature quantity values at the time of work execution

Claims (10)

In the process of assembling the gripping component by the hand provided on the robot arm of the industrial robot, a plurality of fitting / inserting operations associated with the operation ID are repeatedly performed in advance, and the operation quality stored for each operation ID and A robot system for determining the work result of the actual work associated with the work ID and estimating the cause of the failure by organizing the failure causes,
One pre-defined based on a force sensor signal output from a force sensor that detects a force acting on the hand and / or a position detection signal output from each joint axis drive motor provided in the robot arm. Feature quantity extraction means for extracting physical quantities corresponding to the above feature quantity parameters as feature quantities;
For each work ID, history data recording means for recording first history data including each extracted feature quantity, work quality indicating work success or work failure, and cause of work failure;
A feature amount existence region recording unit, which is an optimum feature among the one or more feature amount parameters based on the first history data stored in the history data recording unit and associated with each work ID. A feature set in the coordinate system based on a coordinate system in which a set of feature parameters is selected and each of the selected feature parameter is adopted as a coordinate axis, and the feature amount existing range of the selected optimum feature parameter A feature amount existence region recording means for determining and recording the amount existence region, and
Each feature amount existence area is associated with the work success or failure indicating the work success included in the first history data or the cause of the work failure, and the optimum feature parameter set indicates the work success It is selected based on an overlap area between a feature quantity existence area associated with work quality and each feature quantity existence area associated with the cause of the work failure,
As a result, the work quality determination means determines the work result of the actual work associated with the work ID and estimates the cause of failure based on the feature quantity existence area recorded in the feature quantity existence area recording means. Robot system that makes it possible. The robot system according to claim 1,
The feature quantity existence region based on the feature quantity existence range of the feature quantity parameter is a presence specified in advance based on an average value and variance of the feature quantities of the first history data recorded in the history record data means. A robot system characterized by being calculated with probability. The robot system according to claim 1 or 2,
The work quality determination unit is configured to determine, in the coordinate system, the feature value for the selected optimum feature parameter among the feature values extracted by the feature value extraction unit for the actual work associated with the work ID. The plot is
When it is included only in the feature amount existence area associated with the work quality indicating work success, it is determined as work success,
If it included only in one or more of the said characteristic quantity presence area associated with the cause of the working failure, as well as determination fails and work to extract the failure cause of the feature amount existing region is associated Or
The feature pass / fail that is included in one or more of the feature quantity existence area associated with the work pass / fail indicating the work success and each feature quantity existence area associated with the cause of the work failure, or indicating the work success If it is not included in any of the feature quantity existence area associated with the feature quantity existence area associated with the cause of the work failure, it is determined that the determination is impossible,
A robot system displayed by a work quality / cause display means provided in the robot system. The robot system according to claim 3, wherein
The work pass / fail judgment means further includes the feature quantity existence including the plot when the plot is included only in two or more of the feature quantity existence areas associated with the cause of the work failure. extracts all the two or more failure cause regions are associated, said two or more failure cause is displayed by the working quality - cause display unit, the robot system. The robot system according to claim 3 or 4,
In the case where the work pass / fail determination means determines that the work is successful, the feature quantity existing area associated with the work pass / fail indicating the work success and each feature quantity existing area associated with the cause of the work failure. If there is an overlapping area, based on the size of the overlapping area and the size of the feature amount existing area associated with the work quality indicating work success, calculate the probability of determination when the work is successful,
A robot system configured to further display the accuracy of the determination by the work quality / cause display means. The robot system according to any one of claims 1 to 5,
The one or more feature quantity parameters for the feature quantities extracted by the plurality of feature quantity extraction means are: an insertion amount of a gripping part calculated from the position detection signal; and a peak of time differentiation in the force sensor signal. Any one of a value, an insertion work amount calculated from the insertion amount and the force sensor signal, and a change amount of a hand posture angle calculated from a position detection signal of each joint axis drive motor of the robot arm Robot system, including two or more. The robot system according to any one of claims 1 to 6, further comprising:
The history record data means includes, for the actual work associated with the work ID, the feature quantity extracted by the feature quantity extraction means, the work quality determined by the work quality determination means, and the estimated cause of failure. Configured to additionally record second historical data including,
The feature quantity existence area recording means can update each feature quantity existence area based on the first history data and the second history data stored in the history data recording means. The robot system according to claim 1, further comprising:
The optimal feature quantity parameter set includes a feature quantity existence region associated with the work pass / fail indicating work success from the predetermined number of feature quantity parameter sets among the one or more feature quantity parameters and the work. A robot system in which an overlap region with each feature amount existence region associated with the cause of failure does not exist or a set of feature amount parameters that minimizes the overlap region is selected. The robot system according to claim 8, further comprising:
When there are a plurality of feature quantity parameter sets that do not have the overlapping area, a feature quantity existence area associated with the work quality indicating success of the work and each feature quantity existence area associated with the cause of the work failure The robot system, wherein the optimum feature parameter set is selected based on a distance between centers. The robot system according to claim 8, further comprising:
When there is no feature parameter set that does not include the overlapping region, the predetermined number is incremented, and the feature associated with the work quality indicating the work success is selected from the incremented feature parameter set. A robot system in which an overlap area between a quantity existence area and each feature quantity existence area associated with the cause of the work failure does not exist or a set of feature quantity parameters that minimizes the overlap area is selected.