Interferometric inspection device

Abstract

An interference checking device (1) has: an interference avoidance position learning unit (2) that learns a relationship between a fingertip position and posture of a robot and information related to correction of the fingertip position of the robot for avoiding interference of the robot and a surrounding environment in a case where the fingertip position and posture is specified; and an interference avoidance position calculation unit (3) that, if an action instruction that specifies the fingertip position and posture of the robot is input, calculates information related to correction of the action instruction for avoiding interference of the robot and the surrounding environment based on a learning result obtained by the interference avoidance position learning unit (2).

Description

Technical Field

[0001] This invention relates to an interference inspection device for checking whether or not interference occurs between a robot and its surrounding environment. Background Technology

[0002] In robotic systems, it is required to perform high-speed movements while preventing interference between the robot and its surrounding environment, such as devices within the system. Therefore, when generating motion commands for the robot, an interference check is performed at each point in time to confirm whether the robot, its fore-end hand, or the object held by the hand will collide with the surrounding environment. If interference is predicted, the motion command is corrected.

[0003] For example, the interference inspection device described in Patent Document 1 includes: a model of a multi-joint manipulator and an obstacle created using learning data for a neural network; a neural network that takes the joint angles of the multi-joint manipulator as input and outputs whether the multi-joint manipulator and the obstacle interfere when the joint angles are realized; and a determination unit that compares the output of the neural network with a determination value set by the user and finally determines whether interference has occurred, and the interference inspection device performs interference inspection on the multi-joint manipulator and the obstacle.

[0004] Patent Document 1: Japanese Patent Application Publication No. 7-223181 Summary of the Invention

[0005] Interference checking involves approximating the structural elements of a robotic system, including the robot and its surrounding environment, using shapes like cuboids and spheres, and then verifying whether any of these structural elements interfere with each other. Therefore, interference checking presents the challenge of computational time.

[0006] To address this issue, the interference checking device described in Patent Document 1 uses a neural network that has learned whether the robot and its surroundings interfere with each joint angle of the robot. The interference check is performed through the neural network's calculations, thus reducing the computational load. However, the interference checking device described in Patent Document 1 only determines whether interference exists; to avoid interference, it is necessary to separately determine how to correct the robot's trajectory. Therefore, there is a problem that time is required to generate the trajectory correction when interference occurs.

[0007] The present invention was made in view of the above circumstances, and its object is to provide an interference inspection device capable of determining whether interference occurs between the robot and its surrounding environment and reducing the time required to correct the robot's trajectory in cases where interference occurs.

[0008] To address the aforementioned issues and achieve the objectives, the interference inspection device of the present invention includes an interference avoidance position learning unit that learns the relationship between the robot's fingertip position and orientation and information related to corrections made to the robot's fingertip position to avoid interference with the surrounding environment when that fingertip position and orientation are specified. Furthermore, the interference inspection device includes an interference avoidance position calculation unit that, when inputted with an action command specifying the robot's fingertip position and orientation, calculates information related to corrections made to the action command to avoid interference with the surrounding environment, based on the learning results obtained by the interference avoidance position learning unit.

[0009] The effects of the invention

[0010] The interference inspection device involved in this invention has the following effects: it can determine whether there is interference between the robot and the surrounding environment and shorten the time required to correct the robot's motion trajectory in the event of interference. Attached Figure Description

[0011] Figure 1 This is a diagram illustrating a structural example of the interference inspection device according to Embodiment 1.

[0012] Figure 2 This is a diagram illustrating a structural example of the interference avoidance position learning unit in the interference inspection device according to Embodiment 1.

[0013] Figure 3 This is a flowchart illustrating an example of the operation of the interference avoidance position learning unit in the interference inspection device according to Embodiment 1.

[0014] Figure 4 This is a diagram illustrating a structural example of the interference avoidance position calculation unit in the interference checking device according to Embodiment 1.

[0015] Figure 5 This diagram is used to explain the operation of the operation correction unit of the interference inspection device according to Embodiment 1.

[0016] Figure 6 This is a flowchart illustrating an example of the action of the interference inspection device according to Embodiment 1 in correcting the action command.

[0017] Figure 7 This is a diagram illustrating a structural example of the interference inspection device for the robot according to Embodiment 2.

[0018] Figure 8 This is a diagram illustrating a structural example of the interference avoidance position learning unit in the interference inspection device according to Embodiment 3.

[0019] Figure 9 This is a diagram illustrating an example of the structure of the interference avoidance position learning unit in the interference inspection device according to Embodiment 4.

[0020] Figure 10 This is a diagram used to explain the operation of the interference inspection device according to Embodiment 5.

[0021] Figure 11 This diagram is used to explain the operation of the operation correction unit of the interference inspection device according to Embodiment 7.

[0022] Figure 12 This is a diagram illustrating a structural example of the interference avoidance position calculation unit in the interference checking device according to Embodiment 8.

[0023] Figure 13 This is a diagram illustrating a structural example of an operation command generation device having the interference inspection device according to Embodiment 9.

[0024] Figure 14 This is a diagram illustrating a structural example of an operation command generation device having the interference inspection device according to Embodiment 10. Detailed Implementation

[0025] Hereinafter, based on the accompanying drawings, the interference inspection apparatus according to the embodiments of the present invention will be described in detail.

[0026] Implementation method 1.

[0027] Figure 1 This diagram illustrates a structural example of the interference checking device according to Embodiment 1. The interference checking device 1 according to Embodiment 1 is installed in a computer 100, such as a personal computer. That is, the interference checking device 1 according to Embodiment 1 is implemented by the computer 100 executing a program for operating as the interference checking device 1. The interference checking device 1 includes an interference avoidance position learning unit 2, an interference avoidance position calculation unit 3, an action correction unit 4, and an action planning unit 5.

[0028] The interference checking device 1 checks whether the robot will interfere with its surrounding environment, such as nearby devices, when it moves according to the motion plan generated by the motion planning unit 5 (illustration omitted). In other words, it checks whether the robot will collide with its surroundings. Furthermore, if interference is expected, the interference checking device 1 calculates a correction amount to adjust the robot's position to a position where interference does not occur. Here, the motion plan includes motion instructions that specify the robot's position. Therefore, the interference checking device 1 determines whether the robot will interfere with its surroundings if it moves to the position specified by the motion instruction (i.e., the instruction position) for each motion instruction included in the motion plan. If interference is expected, it calculates a correction amount to adjust the motion instructions in such a way that the motion instructions specify an instruction position where interference does not occur. The determination of whether interference occurs and the calculation of the correction amount in case of interference are performed by the interference avoidance position calculation unit 3. When the interference avoidance position calculation unit 3 detects interference between the robot and its surroundings, the motion correction unit 4 corrects the motion instructions created by the motion planning unit 5 based on the correction amount calculated by the interference avoidance position calculation unit 3.

[0029] The interference avoidance position learning unit 2 learns in advance the relationship between the position and orientation of the robot's fingertips and the vertical upward correction amount required to avoid interference. The position and orientation of the robot's fingertips are specified by the motion instructions generated by the motion planning unit 5, that is, the motion instructions contained in the motion plan. Furthermore, in the following description, "the position and orientation of the robot's fingertips" will sometimes be referred to as "the position and orientation of the robot's fingertips". The result learned by the interference avoidance position learning unit 2, specifically, an approximate function of the relationship between the robot's fingertips position and orientation and the vertical upward correction amount required to avoid interference when the fingertips position and orientation are specified, is transmitted to the interference avoidance position calculation unit 3, which stores the approximate function. The interference avoidance position calculation unit 3 can directly store the function of the neural circuit network, etc., learned by the interference avoidance position learning unit 2, or it can store the parameters of the function of the neural circuit network, etc., learned by the interference avoidance position learning unit 2.

[0030] If motion commands are generated by the motion planning unit 5, the robot's fingertip position posture for each predetermined cycle of motion commands is input to the interference avoidance position calculation unit 3. Furthermore, the input of the robot's fingertip position posture refers to the input of motion commands specifying the robot's fingertip position posture. For each input fingertip position posture of the robot, the interference avoidance position calculation unit 3 determines whether the robot will interfere with the surrounding environment. If interference is determined to occur, it calculates the correction amount (hereinafter referred to as the command correction amount) of the robot's motion commands required to prevent interference and outputs it to the motion correction unit 4. The motion correction unit 4 corrects the motion commands generated by the motion planning unit 5 based on the command correction amount output from the interference avoidance position calculation unit 3. The corrected motion commands from the motion correction unit 4 are sent to the robot control device 200 via the motion planning unit 5. The robot control device 200 implements actual robot motion control based on the corrected motion commands. The robot control device 200 has a command generation unit 6, which creates motion commands to be sent to the robot based on the corrected motion commands.

[0031] Next, the contents of each structural element of the interference inspection device 1 will be explained in detail.

[0032] Figure 2 This diagram illustrates a structural example of the interference avoidance position learning unit 2 included in the interference checking device 1 according to Embodiment 1. The interference avoidance position learning unit 2 includes an interference checking unit 7, a fingertip position correction unit 8, an environmental information storage unit 9, and a learning processing unit 10.

[0033] Figure 2 The interference avoidance position learning unit 2, as shown, first calculates the correction amount required to avoid interference for various fingertip positions and postures of the robot. Specifically, the interference checking unit 7 checks whether the robot will interfere with the surrounding environment for each fingertip position and posture. Here, when the robot has a hand, the interference checking unit 7 checks whether the hand will interfere with the surrounding environment in addition to the robot body itself; when the hand moves in a grasping state, it also checks whether the grasped object will interfere with the surrounding environment. In addition, the environmental information storage unit 9 pre-stores information about the robot body, such as the link length and thickness of the robot arm, the size of the hand, the size of the object, the grasping position of the hand, and the position and size of obstacles in the robot's surrounding environment.

[0034] Based on the various information stored internally in the environmental information storage unit 9, the interference inspection unit 7 approximates each structural element of the robot system, such as the robot and surrounding obstacles, using cuboids, spheres, etc., and checks whether the approximated robot will interfere with all other approximated structural elements. Furthermore, the robot approximation is performed on each structural component of the robot, such as the arm, hand, and the object grasped by the hand.

[0035] The interference checking unit 7 outputs the inspection result to the fingertip position correction unit 8. If the interference checking unit 7 determines that no interference will occur, the fingertip position correction unit 8 sets the fingertip position correction amount of the fingertip position posture to be inspected to 0 and outputs it to the learning processing unit 10. On the other hand, if the interference checking unit 7 determines that interference will occur, the fingertip position correction unit 8 fixes the posture in the fingertip position posture and corrects the fingertip position in the direction specified by the user, outputting the corrected fingertip position posture (hereinafter referred to as the corrected fingertip position posture) to the interference checking unit 7. Furthermore, the interference checking unit 7 checks whether the robot will interfere with the surrounding environment under the corrected fingertip position posture. The corrected fingertip position is set to, for example, the Z direction (the vertical direction of the fingertip position), but the horizontal direction of the fingertip position can also be corrected. Additionally, both the vertical and horizontal directions of the fingertip position can be corrected. Furthermore, in this embodiment, the fingertip position is corrected while the posture is fixed. However, in the case of a vertical 6-axis robot, the fingertip position can also be corrected while fixing the joint positions of the wrist axes (the 4th, 5th, and 6th axes on the fingertip side). The interference detection unit 7 and the fingertip position correction unit 8 repeat the above operation until the interference detection unit 7 determines that there is no interference. When the interference detection unit 7 determines that there is no interference, the fingertip position correction unit 8 outputs the fingertip position correction amount at this time as the fingertip position correction amount of the fingertip position posture set as the inspection object to the learning processing unit 10.

[0036] Next, the learning processing unit 10 learns the approximate function by using a combination of fingertip position posture and fingertip position correction amount as input and output, respectively. Figure 2 The interference avoidance position learning unit 2 shown is used to learn the neural circuit network 20 as an example of an approximation function. The learned neural circuit network 20 is then sent to the interference avoidance position calculation unit 3.

[0037] If the actions of the interference avoidance position learning unit 2 described above are represented by a flowchart, then it becomes... Figure 3 . Figure 3 This is a flowchart illustrating an example of the operation of the interference avoidance position learning unit 2 of the interference inspection device 1 according to Embodiment 1.

[0038] like Figure 3 As shown, the interference avoidance position learning unit 2 first inputs the robot's fingertip position posture (step S11), and the interference checking unit 7 confirms whether interference exists (step S12). That is, in step S12, the interference checking unit 7 confirms whether the robot will interfere with the surrounding environment under the fingertip position posture input in step S11.

[0039] If no interference occurs (step S13: No), the fingertip position correction unit 8 outputs a fingertip position correction value of 0 to the learning processing unit 10, based on the fingertip position correction value input in step S11. Next, the learning processing unit 10 learns the robot's fingertip position posture input in step S11 and the fingertip position correction value input from the fingertip position correction unit 8 (step S15). Specifically, the learning processing unit 10 performs learning of the neural circuit network 20 in such a way that if there is an input of the robot's fingertip position posture, it outputs the fingertip position correction value.

[0040] On the other hand, if interference is likely to occur (step S13: Yes), the fingertip position correction unit 8 corrects the fingertip position of the robot's fingertip position posture (step S14). The fingertip position correction unit 8 outputs the corrected fingertip position posture of the robot (corrected fingertip position posture) to the interference checking unit 7, which confirms whether the robot will interfere with the surrounding environment under the corrected fingertip position posture (step S12). Then, if the interference checking unit 7 determines that no interference will occur (step S13: No), the interference avoidance position learning unit 2 outputs the correction amount of the fingertip position at that moment as the fingertip position correction amount under the fingertip position posture input in step S11 to the learning processing unit 10.

[0041] Interference avoidance position learning unit 2 repeatedly executes various fingertip position poses of the robot as objects. Figure 3 The processing steps S11 to S15 shown are used to learn the neural circuit network 20.

[0042] The interference avoidance position calculation unit 3 inputs the value of the action command generated by the action planning unit 5 for each predetermined cycle. The predetermined cycle is the command generation cycle, and the cycle is a multiple of the command generation cycle.

[0043] Figure 4 This diagram illustrates a structural example of the interference avoidance position calculation unit 3 included in the interference checking device 1 according to Embodiment 1. (See diagram below.) Figure 4As shown, the interference avoidance position calculation unit 3 maintains the trained neural network 20 received from the interference avoidance position learning unit 2. If the robot's fingertip position posture is input from the motion planning unit 5, the interference avoidance position calculation unit 3 inputs the fingertip position posture into the neural network 20 and calculates the fingertip position correction amount under that fingertip position posture. The interference avoidance position calculation unit 3 outputs the calculated fingertip position correction amount and the specified motion command for the robot's fingertip position posture input from the motion planning unit 5 to the motion correction unit 4.

[0044] The motion correction unit 4 corrects the motion command based on the fingertip position correction amount calculated by the interference avoidance position calculation unit 3. Figure 5 This diagram illustrates the operation of the operation correction unit 4 included in the interference checking device 1 according to Embodiment 1. Figure 5 In the diagram, circles (○) represent the values ​​of motion commands output from the motion planning unit 5 for each cycle, and triangles (△) represent the values ​​of the motion commands for each cycle plus the fingertip position correction amount. The track formed by connecting the circles is the fingertip track before correction, and the track formed by connecting the triangles is the fingertip track after correction.

[0045] exist Figure 5 In the example shown, motion commands for seven cycles, including the start and end points of the motion, are output. At three of these points, the fingertip position correction is zero, therefore the motion commands output from the motion planning unit 5 are identical to the corrected motion commands. That is, at these three points, the motion correction unit 4 does not correct the input motion commands. At the remaining four points, the motion correction unit 4 corrects the input motion commands using the fingertip position correction amount output from the interference avoidance position calculation unit 3. The motion correction unit 4 outputs the motion commands corrected for each cycle using the fingertip position correction amount to the motion planning unit 5.

[0046] If the motion planning unit 5 receives a motion command corrected by the motion correction unit 4, it directly outputs it to the command generation unit 6. However, in this embodiment, the motion command corrected by the motion correction unit 4 is re-output to the interference avoidance position calculation unit 3. When the fingertip position correction amount in the interference avoidance position calculation unit 3 becomes 0 for all cycles, the motion planning unit 5 sets the correction in the motion correction unit 4 to complete and outputs the motion command at this time to the command generation unit 6 of the robot control device 200.

[0047] If the above actions—that is, the actions by which the interference checking device 1 corrects the action commands so that the robot does not interfere with the surrounding environment—are represented by a flowchart, then it becomes... Figure 6 . Figure 6This is a flowchart illustrating an example of how the interference checking device 1, according to Embodiment 1, corrects an action command.

[0048] First, the robot's fingertip position and posture are input from the motion planning unit 5 to the interference avoidance position calculation unit 3 (step S21). The interference avoidance position calculation unit 3 calculates the fingertip position correction amount based on the input fingertip position and posture (step S22). The interference avoidance position calculation unit 3 outputs the calculated fingertip position correction amount together with the input motion command specifying the fingertip position and posture, i.e., the robot's fingertip position and posture, to the motion correction unit 4.

[0049] Next, the motion correction unit 4 confirms the fingertip position correction amount calculated by the interference avoidance position calculation unit 3. If the fingertip position correction amount is 0 (step S23: Yes), the motion planning unit 5 outputs the motion command to the command generation unit 6 of the robot control device 200 (step S26). That is, in step S26, the motion correction unit 4 outputs the motion command received from the interference avoidance position calculation unit 3 along with the fingertip position correction amount directly to the motion planning unit 5 without correction. At this time, the motion correction unit 4 notifies the motion planning unit 5 that the motion command will not be corrected. The motion planning unit 5 then outputs the motion command received from the motion correction unit 4 to the command generation unit 6.

[0050] In this case, the instruction generation unit 6 of the robot control device 200 generates an action instruction for each control cycle of the robot based on the action instructions input from the action planning unit 5. The action instruction specifies a trajectory that smoothly connects the locations specified by each input action instruction.

[0051] On the other hand, when the fingertip position correction amount is not zero (step S23: No), the motion correction unit 4 corrects the robot's fingertip position based on the fingertip position correction amount calculated by the interference avoidance position calculation unit 3 (step S24). Furthermore, the correction of the robot's fingertip position by the motion correction unit 4 refers to the correction of the motion command received from the interference avoidance position calculation unit 3 along with the fingertip position correction amount. Next, the fingertip position posture corrected by the motion correction unit 4 is input to the interference avoidance position calculation unit 3 (step S25). In step S25, the motion correction unit 4 outputs the corrected motion command specifying the robot's fingertip position posture to the motion planning unit 5. At this time, the motion correction unit 4 notifies the motion planning unit 5 that the motion command has been corrected. The motion planning unit 5 outputs the motion command received from the motion correction unit 4 to the interference avoidance position calculation unit 3.

[0052] The interference avoidance position calculation unit 3, the action correction unit 4, and the action planning unit 5 of the interference inspection device 1 repeat the above steps S21 to S26 until the fingertip position correction amount calculated by the interference avoidance position calculation unit 3 becomes 0.

[0053] As described above, the interference checking device 1 of Embodiment 1 pre-learns various fingertip positions and postures of the robot and the correction amount of the fingertip position used to prevent the robot from interfering with the surrounding environment through the interference avoidance position learning unit 2. When outputting motion commands to the robot control device 200, the interference avoidance position calculation unit 3 calculates the fingertip position correction amount used to correct the motion commands using the learning results in the interference avoidance position learning unit 2, so as to specify the fingertip position and posture that does not interfere. According to the interference checking device 1, the presence or absence of interference is checked, which can reduce the amount of calculation required to correct motion commands when interference exists, and as a result, the time required to correct motion commands can be shortened.

[0054] Implementation method 2.

[0055] Figure 7 This diagram illustrates a structural example of the interference checking device for the robot according to Embodiment 2. The difference between the interference checking device 1a according to Embodiment 2 and the interference checking device 1 according to Embodiment 1 is that the robot control device 200 includes an interference avoidance position calculation unit 3, an action planning unit 5, and an action correction unit 4. Furthermore, regarding the overall operation, instead of the action correction unit 4 repeatedly calculating the fingertip position correction amount by the interference avoidance position calculation unit 3 to correct the specified action command for the robot's fingertip position posture, the action command initially corrected by the action correction unit 4 is directly output to the command generation unit 6. Aside from these differences, it is the same as Embodiment 1, and therefore, the description is omitted. Moreover, the action command initially corrected by the action correction unit 4 is output to the command generation unit 6 via the action planning unit 5, just as in Embodiment 1.

[0056] The interference checking device 1a according to Embodiment 2 is configured to calculate the fingertip position correction amount for avoiding interference and correct the motion command inside the robot control device 200. Therefore, the motion command that takes interference avoidance into account is corrected immediately each time the motion plan is made, which enables the robot to move.

[0057] Implementation method 3.

[0058] Next, the interference checking device according to Embodiment 3 will be described. The structure of the interference checking device according to Embodiment 3 is the same as that of Embodiment 1 (see [reference]). Figure 1Furthermore, in the following description, for convenience, the interference checking device according to Embodiment 3 will be referred to as interference checking device 1b. In this embodiment, the parts that differ from Embodiment 1 will be described.

[0059] The difference between the interference inspection device 1b in Embodiment 3 and Embodiment 1 is that, in the interference avoidance position learning unit 2 and the interference avoidance position calculation unit 3, the fingertip position correction amount is not used as the output of the neural circuit network 20, but the corrected fingertip position posture is used instead.

[0060] Figure 8 This diagram illustrates a structural example of the interference avoidance position learning unit 2b in the interference checking device 1b according to Embodiment 3. The interference avoidance position learning unit 2b includes an interference checking unit 7, a fingertip position correction unit 8b, an environmental information storage unit 9, and a learning processing unit 10b. The interference checking unit 7 and the environmental information storage unit 9 are the same as those in the interference avoidance position learning unit 2 according to Embodiment 1, and therefore their description is omitted.

[0061] like Figure 8 As shown, the fingertip position correction unit 8b does not output the fingertip position correction amount described in Embodiment 1, but rather outputs the corrected fingertip position posture. In the learning processing unit 10b, the neural circuit network 20 performs learning by using the combination of the robot's fingertip position posture and the corrected fingertip position posture as input and output, respectively.

[0062] The interference inspection device 1b according to embodiment 3 uses the learning results in the interference avoidance position learning unit 2b to calculate the fingertip position posture that can avoid interference at each location specified by the action command by the interference avoidance position calculation unit 3. Therefore, the amount of calculation required to calculate the fingertip position posture that does not cause interference can be reduced.

[0063] Implementation method 4.

[0064] Next, the interference checking device according to Embodiment 4 will be described. The structure of the interference checking device according to Embodiment 4 is the same as that of Embodiment 1 (see [reference]). Figure 1 Furthermore, in the following description, for convenience, the interference checking device according to Embodiment 4 will be referred to as interference checking device 1c. In this embodiment, the parts that differ from Embodiment 1 will be described.

[0065] The difference between the interference inspection device 1c in Embodiment 4 and Embodiment 1 is that, in the interference avoidance position learning unit 2 and the interference avoidance position calculation unit 3, the fingertip position posture is not used as the input to the neural circuit network 20, but the joint positions of each axis of the robot are used instead.

[0066] Figure 9 This diagram illustrates a structural example of the interference avoidance position learning unit 2c included in the interference checking device 1c according to Embodiment 4. The interference avoidance position learning unit 2c includes an interference checking unit 7c, a fingertip position correction unit 8, an environmental information storage unit 9, and a learning processing unit 10c. The fingertip position correction unit 8 and the environmental information storage unit 9 are identical to those in the interference avoidance position learning unit 2 according to Embodiment 1, and therefore their description is omitted.

[0067] The interference checking unit 7c checks whether there is interference between the robot and its surrounding environment based on the joint position of the robot specified by the input motion command and the information stored in the environmental information storage unit 9. The operation of the interference checking unit 7c to check for the presence or absence of interference is the same as that of the interference checking unit 7 in Embodiment 1, except that it uses the joint position of the robot instead of the fingertip position posture.

[0068] exist Figure 9 In the interference avoidance position learning unit 2c shown, the fingertip position correction unit 8 outputs the fingertip position correction amount, but it can also output it after converting the fingertip position correction amount into a joint position correction amount. When the joint position correction amount is learned by the interference avoidance position learning unit 2c, the joint position correction amount is also output in the interference avoidance position calculation unit 3.

[0069] The interference checking device 1c according to Embodiment 4 uses the joint positions of the robot as input to the interference avoidance position learning unit 2c and the interference avoidance position calculation unit 3. Therefore, when the motion planning unit 5 generates instructions to make each joint move synchronously, the amount of computation for calculating instructions that can avoid interference can be reduced.

[0070] Implementation method 5.

[0071] Next, the interference checking device according to Embodiment 5 will be described. The structure of the interference checking device according to Embodiment 5 is the same as that of Embodiment 1 (see [reference]). Figure 1 Furthermore, in the following description, for convenience, the interference checking device according to Embodiment 5 will be referred to as interference checking device 1d. In this embodiment, the parts that differ from Embodiment 1 will be described.

[0072] The difference between the interference checking device 1d in Embodiment 5 and Embodiment 1 lies in the method for correcting the motion command in the motion correction unit 4. In the following description, the motion correction unit 4 in Embodiment 5 will be referred to as motion correction unit 4d.

[0073] In the motion correction unit 4 of the interference detection device 1 according to Embodiment 1, the motion command is corrected in such a way as smoothly connecting the corrected fingertip position posture of each cycle. However, in the motion correction unit 4d of the interference detection device 1d according to Embodiment 5, the motion command determined based on the fingertip position correction amount calculated by the interference avoidance position calculation unit 3 is combined with the motion command generated by the motion planning unit 5 to generate a new motion command. Figure 10 This diagram illustrates the operation of the interference checking device 1d according to Embodiment 5. The motion correction unit 4d calculates a corrected motion command, which is a combination of the motion command generated by the motion planning unit 5 and the fingertip position correction amount output by the interference avoidance position calculation unit 3. Figure 10 As shown, the calculated corrected motion command and the motion command generated by the motion planning unit 5 are combined. Furthermore, the result of the combination by the motion correction unit 4d is repeatedly input into the interference avoidance position calculation unit 3, and the motion correction unit 4d recalculates the corrected motion command according to the fingertip position correction amount calculated by the interference avoidance position calculation unit 3. This series of processes is repeated until the fingertip position correction amount calculated by the interference avoidance position calculation unit 3 becomes 0.

[0074] After the fingertip position correction amount calculated by the interference avoidance position calculation unit 3 becomes 0, the motion planning unit 5 outputs the corrected motion command calculated by the motion correction unit 4d and the motion command initially generated by the motion planning unit 5 to the command generation unit 6. The command generation unit 6 synthesizes the motion command initially generated by the motion planning unit 5 and the corrected motion command generated by the motion correction unit 4d to generate the robot's motion command.

[0075] The interference checking device 1d according to Embodiment 5 can generate instructions that can perform actions at high speed while reducing the amount of calculation required for action plans to avoid interference by synthesizing action instructions generated without considering interference and corrective action instructions calculated with consideration of interference.

[0076] Implementation method 6.

[0077] Next, the interference checking device according to Embodiment 6 will be described. The structure of the interference checking device according to Embodiment 6 is the same as that of Embodiment 1 (see Figure 1). Figure 1 Furthermore, in the following description, for convenience, the interference checking device according to Embodiment 6 will be referred to as interference checking device 1e. In this embodiment, the parts that differ from Embodiment 1 will be described.

[0078] The interference checking device 1 of Embodiment 1 checks for interference and corrects the fingertip position for each predetermined cycle. In contrast, the interference checking device 1e of Embodiment 6 performs interference checking and calculates the fingertip position correction amount for each predetermined cycle. However, when generating motion commands through the motion planning unit 5, it specifies several transit points and generates motion commands that do not cause interference by changing these transit points in the motion correction unit 4. Furthermore, the motion correction unit 4 outputs the transit points where all fingertip position correction amounts calculated by the interference avoidance position calculation unit 3 become 0, along with the motion commands generated by the motion planning unit 5, to the command generation unit 6 via the motion planning unit 5. The command generation unit 6 generates commands for the robot based on the input motion commands and transit points, with the robot's motion trajectory passing through each input transit point.

[0079] When the interference checking device 1e according to embodiment 6 specifies the transit point during the action planning, it can reduce the amount of computation required to generate action instructions that avoid interference.

[0080] Implementation method 7.

[0081] Next, the interference checking apparatus according to Embodiment 7 will be described. The structure of the interference checking apparatus according to Embodiment 7 is the same as that of Embodiment 1 (see [reference]). Figure 1 Furthermore, in the following description, for convenience, the interference checking device according to Embodiment 7 will be referred to as interference checking device 1f. In this embodiment, the parts that differ from Embodiment 1 will be described.

[0082] In the interference inspection device 1 according to Embodiment 1, the fingertip position correction amount for avoiding interference is set as the output of the neural circuit network 20 in the interference avoidance position learning unit 2 and the interference avoidance position calculation unit 3. However, in the interference inspection device 1f according to this embodiment, the upper limit and lower limit values ​​of the fingertip position where no interference occurs are set as the output.

[0083] Figure 11 This diagram illustrates the operation of the operation correction unit 4 included in the interference inspection device 1f according to Embodiment 7. The operation correction unit 4 of the interference inspection device 1f is as follows: Figure 11 As shown, the range of fingertip positions that do not interfere (the range of non-interference) is determined based on the upper and lower limits of the fingertip positions in each cycle. The trajectory of the fingertip position is determined by keeping the fingertip position within the determined range, thereby calculating the motion command correction value that avoids interference. The motion command correction value calculated by the motion correction unit 4 is output to the command generation unit 6 via the motion planning unit 5.

[0084] The interference checking device 1f according to Embodiment 7 can optimize motion commands within the range of the robot's fingertip position without interference, thus reducing the time required to generate motion commands for the robot that achieves high-speed motion while avoiding interference.

[0085] Implementation method 8.

[0086] Next, the interference checking apparatus according to Embodiment 8 will be described. The structure of the interference checking apparatus according to Embodiment 8 is the same as that of Embodiment 1 (see...). Figure 1 Furthermore, in the following description, for convenience, the interference checking device according to Embodiment 8 will be referred to as interference checking device 1g. In this embodiment, the parts that differ from Embodiment 1 will be described.

[0087] In the interference inspection device 1 according to Embodiment 1, the neural network 20 used in the interference avoidance position learning unit 2 and the interference avoidance position calculation unit 3 is singular and independent of the robot's range of motion. In contrast, in the interference inspection device 1g according to Embodiment 8, the robot's range of motion is divided into multiple regions, and the interference avoidance position learning unit 2 performs learning for each region.

[0088] Figure 12 This is a diagram illustrating a structural example of the interference avoidance position calculation unit 3g included in the interference checking device 1g according to Embodiment 8. (See diagram below.) Figure 12 As shown, the interference avoidance position calculation unit 3g has a region selection unit 11 and a correction amount determination unit 12. In addition, it maintains the multiple trained neural circuits 20 received from the interference avoidance position learning unit 2.

[0089] The region selection unit 11 determines which of multiple regions the robot's fingertip position posture input to the interference avoidance position calculation unit 3g corresponds to, and calculates the fingertip position correction amount using the neural circuit network 20 corresponding to the determined region. The correction amount determination unit 12 determines the fingertip position correction amount output from the interference avoidance position calculation unit 3g to the motion correction unit 4 based on the fingertip position correction amounts output from each of the neural circuit networks 20. The correction amount determination unit 12 can directly output the output from the neural circuit network 20 corresponding to the region selected by the region selection unit 11. Furthermore, when the input fingertip position posture is near the boundary of a region, the correction amount determination unit 12 performs calculations separately on the outputs from the neural circuit network 20 corresponding to the region to which the input fingertip position posture belongs and the neural circuit network 20 corresponding to the nearby region, and can use the weighted sum of the outputs of these multiple neural circuit networks 20 as the fingertip position correction amount output to the motion correction unit 4. Furthermore, when the region to which the fingertip position posture belongs in the previous cycle is different from the region to which the fingertip position posture belongs in the current cycle, the correction amount determination unit 12 performs calculations on the outputs from the neural circuits 20 corresponding to the regions to which the fingertip position posture belongs in the previous cycle and the regions to which the fingertip position posture belongs in the current cycle, and can use the weighted sum of the outputs of these two neural circuits 20 as the fingertip position correction amount output to the motion correction unit 4.

[0090] The structure and operation of the interference avoidance position calculation unit 3g have been explained. However, when the interference avoidance position learning unit 2 of the interference inspection device 1g performs learning, it is divided into multiple regions, and neural circuit networks 20 are learned in each of the multiple regions, generating neural circuit networks 20 corresponding to each of the multiple regions. That is, when learning is performed by setting the regions existing in the multiple regions as objects, the fingertip position and posture of the robot input to the interference avoidance position learning unit 2 are changed into various values ​​in the regions of the learning objects, and the interference avoidance position learning unit 2 performs learning.

[0091] The interference checking device 1g according to Embodiment 8 can reduce the size of each neural circuit network 20 by dividing the fingertip position and posture of the robot into multiple regions, thereby reducing the computational load required for learning and inference processing of the neural circuit network 20. That is, it can reduce the computational load required in the interference avoidance position learning unit and the interference avoidance position calculation unit.

[0092] Furthermore, the structure of dividing the robot's range of motion into multiple regions for processing in the interference inspection device 1 according to Embodiment 1 has been described, but similarly, the interference inspection devices according to Embodiments 2 to 7 can also divide the robot's range of motion into multiple regions for processing.

[0093] Implementation method 9.

[0094] Next, the interference checking apparatus according to Embodiment 9 will be described. An example of using the interference checking apparatus 1h according to Embodiment 9 will be presented below. Figure 13 As shown. Figure 13 This diagram illustrates a structural example of the motion command generation apparatus having the interference checking device 1h according to Embodiment 9. The motion command generation apparatus is implemented by a computer 100 and includes the interference checking device 1h, a motion correction unit 4, a motion planning unit 5, a passpoint candidate calculation unit 31, and a passpoint candidate determination unit 32. The interference checking device 1h is composed of an interference avoidance position learning unit 2 and an interference avoidance position calculation unit 3. These units are the same as the interference avoidance position learning unit 2 and interference avoidance position calculation unit 3 constituting the interference checking device 1 according to Embodiment 1, therefore, their description is omitted.

[0095] In the computer 100 equipped with the interference checking device 1h, an action plan from the start point to the end point of the action is implemented, generating instructions for the robot. Therefore, the robot body determines the passage points where the hand attached to the robot and the object grasped by the hand will not collide with the surrounding environment. When determining the passage points, firstly, the passage point candidate calculation unit 31 calculates the candidate passage points (hereinafter, sometimes referred to as passage point candidates). The passage point candidate calculation unit 31 inputs the calculated fingertip position posture of the robot at the passage points (candidate passage points) to the interference avoidance position calculation unit 3 inside the interference checking device 1h. In the interference avoidance position calculation unit 3, the fingertip position correction amount is calculated based on the input information of the robot's fingertip position posture. The calculated fingertip position correction amount is sent to the passage point candidate determination unit 32. The passage point candidate determination unit 32 also inputs the robot's fingertip position posture of the candidate passage points calculated by the passage point candidate calculation unit 31. The passage point candidate determination unit 32 determines the robot's fingertip position and posture at the passing point based on the robot's fingertip position and posture input from the passing point candidate calculation unit 31 and the fingertip position correction amount input from the interference avoidance position calculation unit 3. The robot's fingertip position and posture at the passing point determined by the passing point candidate determination unit 32 is output to the motion planning unit 5. The passing point calculated and determined by the passing point candidate calculation unit 31 and the passing point candidate determination unit 32 can be set to one or more points between the start point and the end point of the motion.

[0096] The motion planning unit 5 generates motion commands that start from the motion start point, sequentially pass through the points determined by the pass-through point candidate determination unit 32, and reach the motion end point. It then inputs the robot's fingertip position and posture for each predetermined cycle of motion commands to the interference avoidance position calculation unit 3. Hereinafter, the motion planning unit 5, motion correction unit 4, and interference avoidance position calculation unit 3 perform the same actions as in Embodiment 1. Descriptions of these actions are omitted.

[0097] The interference checking device according to Embodiment 9 has the effect that it can reduce the amount of calculation required to derive the passing points based on the action plan.

[0098] Implementation method 10.

[0099] Next, the interference checking apparatus according to Embodiment 10 will be described. An example of the use of the interference checking apparatus 1i according to Embodiment 10 will be given below. Figure 14 As shown. Figure 14 This diagram illustrates a structural example of an action command generation device having the interference checking device 1i according to Embodiment 10. The action command generation device is implemented by a computer 100 and includes the interference checking device 1i, an initial action path setting unit 33, a passable area determination unit 34, and an optimal action exploration unit 35. The interference checking device 1i is composed of an interference avoidance position learning unit 2 and an interference avoidance position calculation unit 3. These units are the same as the interference avoidance position learning unit 2 and interference avoidance position calculation unit 3 constituting the interference checking device 1f according to Embodiment 7, therefore, their description is omitted.

[0100] The initial motion path setting unit 33 generates multiple paths connecting the start and end points of the motion. The initial motion path setting unit 33 further divides each generated path using a predetermined number of divisions, and outputs the robot's fingertip position and posture at each of the divided locations to the interference avoidance position calculation unit 3 and the passable area determination unit 34. The interference avoidance position calculation unit 3 calculates the upper and lower limits of the fingertip position that will not cause interference based on the robot's fingertip position and posture input from the initial motion path setting unit 33, and outputs this calculation to the passable area determination unit 34. The passable area determination unit 34 calculates the range of the robot's fingertip position and posture that will not cause interference based on the robot's fingertip position and posture input from the initial motion path setting unit 33 and the upper and lower limits of the fingertip position that will not cause interference input from the interference avoidance position calculation unit 33, and defines this range as the passable area.

[0101] The passable area calculated by the passable area determination unit 34 is output to the optimal motion exploration unit 35. The optimal motion exploration unit 35 explores motion commands that connect the start point and end point of the motion, aiming to shorten the motion time while avoiding interference. When exploring motion commands, the optimal motion exploration unit 35 explores the condition that the fingertip position of the explored motion command always remains within the passable area input from the passable area determination unit 34. The motion commands obtained by the exploration performed by the optimal motion exploration unit 35 are output to the command generation unit 6 of the robot control device 200.

[0102] When the interference checking device according to Embodiment 10 explores the action that minimizes the action time under the constraint of avoiding interference, it does not need to perform interference checking calculations every time, and thus has the effect of reducing the amount of calculation required for action exploration.

[0103] The structure shown in the above embodiments is an example, and it can also be combined with other known technologies, and the embodiments can be combined with each other. Without departing from the spirit of the subject, some parts of the structure can be omitted or changed.

[0104] Explanation of the label

[0105] 1. Interference checking device (1a, 1h, 1i); 2. Interference avoidance position learning unit (2b, 2c); 3. Interference avoidance position calculation unit (3g); 4. Motion correction unit (4); 5. Motion planning unit (5); 6. Instruction generation unit (6); 7. Interference checking unit (7c); 8. Fingertip position correction unit (8b); 9. Environmental information storage unit (9); 10. Learning processing unit (10b, 10c); 11. Region selection unit (11); 12. Correction amount determination unit (12); 20. Neural circuit network (20); 31. Passing point candidate calculation unit (31); 32. Passing point candidate determination unit (32); 33. Initial motion path setting unit (33); 34. Passable region determination unit (34); 35. Optimal motion exploration unit (35); 100. Computer (200); Robot control device (200).

Claims

1. An interference detection device, characterized in that, have: The interference avoidance position learning unit learns the relationship between the robot's fingertip position and attitude and information related to the correction of the robot's fingertip position to avoid interference between the robot and the surrounding environment, or the relationship between the robot's joint position and information related to the correction of the robot's joint position to avoid interference between the robot and the surrounding environment. as well as The interference avoidance position calculation unit, when inputted with a motion command specifying the fingertip position and posture of the robot, calculates information related to the correction of the motion command to avoid interference between the robot and the surrounding environment, based on the learning results obtained by the interference avoidance position learning unit. The information calculated by the interference avoidance position calculation unit is used as information related to the correction of the robot's fingertip position or information related to the correction of the robot's joint position. The interference checking device also has: The motion planning unit generates motion commands that specify the fingertip position and posture of the robot and outputs them to the interference avoidance position calculation unit. as well as The motion correction unit corrects the motion command generated by the motion planning unit based on the information calculated by the interference avoidance position calculation unit. The motion correction unit calculates a corrected motion command synthesized with the motion command based on the information calculated by the interference avoidance position calculation unit, and synthesizes the calculated corrected motion command and the motion command to generate a motion command that specifies the corrected fingertip position and posture of the robot.

2. The interference inspection device according to claim 1, characterized in that, The information calculated by the interference avoidance position calculation unit is set as the correction amount for the fingertip position of the robot.

3. The interference inspection device according to claim 2, characterized in that, The correction amount for the fingertip position of the robot represents either or both of the correction amount in the vertical direction and the horizontal direction of the fingertip position.

4. The interference inspection device according to claim 2, characterized in that, The correction amount for the fingertip position of the robot represents either or both of the vertical correction amount and the horizontal correction amount for the fingertip position when the displacement of the robot's wrist axis is fixed.

5. The interference inspection apparatus according to any one of claims 2 to 4, characterized in that, The position and posture of the robot's fingertips are represented by the displacement of each joint of the robot.

6. The interference inspection device according to claim 1, characterized in that, The information calculated by the interference avoidance position calculation unit is set as the corrected fingertip position and posture information of the robot.

7. The interference inspection apparatus according to any one of claims 2 to 5, characterized in that, The interference avoidance position learning unit learns the upper and lower limits of the correction amount for the fingertip position of the robot to prevent the robot from interfering with the surrounding environment.

8. The interference inspection device according to claim 7, characterized in that, have: The motion planning unit generates motion commands that specify the fingertip position and posture of the robot and outputs them to the interference avoidance position calculation unit. as well as The motion correction unit corrects the motion command generated by the motion planning unit based on the upper and lower limits of the correction amount calculated by the interference avoidance position calculation unit, so that the position of the robot's fingertip is between the upper and lower limits.

9. The interference inspection device according to claim 1, characterized in that, Based on the motion command, the motion planning unit generates the transit point of the robot's fingertip position. The motion correction unit, based on the information calculated by the interference avoidance position calculation unit, corrects the transit point instead of the motion command, thereby avoiding interference between the robot and the surrounding environment.

10. The interference inspection apparatus according to any one of claims 1 to 9, characterized in that, The interference-avoidance position learning unit divides the robot's motion region into multiple regions, and performs the learning in each of the multiple regions. The interference avoidance position calculation unit calculates the information related to the correction of the action command for each of the multiple regions based on the learning results obtained by the interference avoidance position learning unit.

Images