Robot control system, robot control method, robot control program and robot

Abstract

To provide a robot control system in which a robot apologizes in an appropriate bowing motion according to severity.SOLUTION: A robot control system (10) includes a robot (12) capable of a bowing motion and, when an operation terminal (16) or a robot (12) determines that the robot has failed, it determines severity in all the categories of the failure, such as temporal loss, pecuniary loss and injuries (physical loss). The robot (12) calculates time spent until it bows (a leaning forward time tb), an angle of a bow (a leaning forward angle θb), and time for bowing (a leaning forward duration time tk) according to the determined severity, and performs a series of bowing motions by controlling an actuator according to the calculated leaning forward time (tb), leaning forward angle (θb), leaning forward duration time (tk) and a time for returning to the original position (a restoration time te).SELECTED DRAWING: Figure 5

Description

[Technical Field]

[0001] The present invention relates to a robot control system, a robot control method, a robot control program, and a robot, and more particularly to a robot control system, a robot control method, a robot control program, and a robot in which a robot or an agent apologizes with a bowing motion, for example. [Background technology]

[0002] In Non-Patent Document 1, the present inventors have revealed that a mitigating effect on failure can be expected through the reactive behavior and non-verbal behavior of a robot, such as facial expressions.

[0003] Furthermore, Non-Patent Document 2 confirms that apologies from multiple robots are more accepted and trusted than apologies from a single robot. [Prior art documents] [Non-patent literature]

[0004] [Non-Patent Document 1] M. Shiomi, K. Nakagawa, and N. Hagita, “Design of a gaze behavior at a small mistake moment for a robot,” Interaction Studies, vol. 14, no. 3, pp. 317-328, 2013. [Non-patent document 2] Y. Okada, M. Kimoto, T. Iio, K. Shimohara, and M. Shiomi, “Two is better than one: Apologies from two robots are preferred,” Plos one, vol. 18, no. 2, pp. e0281604, 2023. DISCLOSURE OF THE INVENTION [Problem to be solved by the invention]

[0005] Non-Patent Documents 1 and 2 demonstrated the effectiveness of apologizing behavior of social robots in response to failure, but did not pay sufficient attention to bowing behavior that corresponds to the seriousness of the failure. In other words, there was no technology that could generate an appropriate apology behavior for a robot depending on the seriousness of the failure.

[0006] Therefore, a primary object of the present invention is to provide a novel robot control system, a novel robot control method, a novel robot control program, and a novel robot.

[0007] Another object of the present invention is to provide a robot control system, a robot control method, a robot control program, and a robot in which a robot or an agent apologizes with an appropriate bowing motion according to the seriousness of the failure. [Means for solving the problem]

[0008] In order to solve the above problems, the present invention employs the following configuration: Note that the reference numerals and supplementary explanations in parentheses indicate the correspondence with the embodiments described later to aid in understanding the present invention, and do not limit the present invention in any way.

[0009] The first embodiment is a robot control system for controlling a robot capable of bowing by leaning at least one of its upper body and head forward, and includes a seriousness determination unit that determines the seriousness of the category of failure when the robot makes a mistake, a parameter value calculation unit that calculates a parameter value that controls the bowing movement based on the seriousness determined by the seriousness determination unit, and a movement control unit that controls the bowing movement of the robot in accordance with the forward lean parameter value calculated by the parameter value calculation unit.

[0010] In the first embodiment, a robot control system (10: reference numerals illustrating corresponding parts in the embodiments, not intended to be limiting; the same applies below) controls a robot (12) capable of bowing by leaning at least one of its upper body and head forward. When the robot fails, a seriousness determination unit (56, 76e, 22; 40, 84d1, S02) determines the seriousness of each failure category, such as time loss, monetary loss, or physical loss. Whether the robot has failed may be determined by the operation terminal (16) or by the robot (12) itself. A parameter value calculation unit (40, 84d1, S11-S13) calculates parameter values ​​(tb, θb, tk, and te) for controlling the bowing motion based on the seriousness determined by the seriousness determination unit. A motion control unit (40, 84d2, S21-S23) controls the bowing motion of the robot in accordance with the forward lean parameter value calculated by the parameter value calculation unit.

[0011] According to the first embodiment, an apology is made with a bowing motion that corresponds to the seriousness of the apology, which makes the apology more easily accepted.

[0012] The second embodiment is a robot control system subordinate to the first embodiment, in which the parameter value calculation unit includes a forward lean angle calculation unit that calculates the forward lean angle of the robot based on the severity, and the movement control unit controls the bowing movement of the robot in accordance with the forward lean angle calculated by the forward lean angle calculation unit.

[0013] In the second embodiment, a forward lean angle calculation unit (40, S11) included in the parameter value calculation unit calculates the forward lean angle (θb) of the robot (12) based on the severity, for example, according to equation (1) described below, and the movement control unit controls the robot to perform a bowing movement in accordance with the forward lean angle.

[0014] According to the second embodiment, by controlling the bowing action of the robot using the forward lean angle calculated based on the seriousness, it is possible to express an apology appropriately according to the seriousness.

[0015] A third embodiment is a robot control system dependent on the second embodiment, in which the parameter value calculation unit further includes a forward lean duration calculation unit that calculates a forward lean duration for which the robot should maintain a forward lean angle based on the severity, and the movement control unit controls the bowing movement in accordance with the forward lean duration calculated by the forward lean duration calculation unit.

[0016] In the third embodiment, a forward lean duration calculation unit (40, S13) included in the parameter value calculation unit calculates the forward lean duration (tk) for which the robot should maintain the forward lean angle based on the severity, and the movement control unit controls the bowing movement in accordance with the forward lean duration.

[0017] According to the third embodiment, the duration of the bow, i.e., the duration of the forward lean, is the easiest way to express the seriousness of the situation. Therefore, by controlling the bowing action of the robot using the forward lean angle and forward lean duration calculated based on the seriousness of the situation, an apology can be expressed more appropriately according to the seriousness of the situation.

[0018] A fourth embodiment is a robot control system dependent on the first or second embodiment, in which the parameter value calculation unit further includes a forward lean time calculation unit that calculates the time until the robot leans forward based on the severity, and the movement control unit controls the bowing movement in accordance with the forward lean time calculated by the forward lean time calculation unit.

[0019] In the fourth embodiment, a forward lean time calculation unit (40, S12) included in the parameter value calculation unit calculates the time it takes for the robot to perform a forward leaning motion, i.e., the forward lean speed (tb), based on the severity, and the motion control unit controls the bowing motion according to the forward lean speed.

[0020] According to the fourth embodiment, the seriousness is further determined by the time it takes to bow, i.e., the forward leaning speed. Therefore, by controlling the robot's bowing action using the forward leaning angle and forward leaning duration at the forward leaning speed calculated based on the seriousness, an apology can be expressed more appropriately according to the seriousness.

[0021] The fifth embodiment is a robot control system subordinate to the first embodiment, further comprising a judgment unit that judges whether an apology accompanied by a bowing action has been accepted, and a parameter value change unit that, when the judgment unit determines that the apology has not been accepted, changes the parameter value to increase it in the same failure category, and the action control unit performs the bowing action again in accordance with the parameter value changed by the parameter value change unit.

[0022] In the fifth example, the determination unit (56, 40, S9) determines whether an apology accompanied by a bowing motion has been accepted. This determination is made, for example, based on video from the camera (50) of the robot (12) and / or audio from the microphone (54), for example, based on video and / or audio of the user (other person) input from the camera 50. For example, it determines whether the other person's angry face has changed to a smile, or whether they have accepted or made a positive statement in response to the apology. Then, when the determination unit determines that the apology has not been accepted, the parameter value change unit (56, 40, S10) changes the parameter value to increase it for the same failure category, and the operation control unit performs the bowing motion again in accordance with the parameter value changed by the parameter value change unit.

[0023] According to the fifth embodiment, even if an apology is not accepted, by increasing the seriousness and performing the bowing action again, the apology may eventually be accepted.

[0024] The sixth embodiment is a method for controlling a robot that can bow by tilting at least one of its upper body and head forward, in which, when the robot makes a mistake, the severity of the category of the mistake is determined, a parameter value for controlling the bowing movement is calculated based on the determined severity, and the bowing movement is controlled in accordance with the calculated parameter value.

[0025] A seventh example is a robot control program for controlling a robot capable of bowing by leaning at least one of its upper body and head forward, the robot control program causing the robot's computer to function as a seriousness determination unit that determines the seriousness of the category of failure when the robot fails, a parameter value calculation unit that calculates a parameter value that controls the bowing motion based on the seriousness determined by the seriousness determination unit, and a motion control unit that controls the robot's bowing motion in accordance with the forward lean parameter value calculated by the parameter value calculation unit.

[0026] The eighth embodiment is a robot capable of bowing by leaning at least one of its upper body and head forward, and is equipped with a seriousness determination unit that determines the seriousness of the category of failure when the robot makes a mistake, a parameter value calculation unit that calculates a parameter value that controls the bowing movement based on the seriousness determined by the seriousness determination unit, and a movement control unit that controls the bowing movement of the robot in accordance with the forward lean parameter value calculated by the parameter value calculation unit.

[0027] Any of the sixth to eighth embodiments can be expected to provide the same effects as the first embodiment. [Effects of the Invention]

[0028] According to this invention, the robot or agent apologizes with an appropriate bowing motion according to the seriousness of the mistake, making it possible to realize an apology that is easily accepted by people.

[0029] The above and other objects, features and advantages of the present invention will become more apparent from the following detailed description of the preferred embodiments with reference to the drawings. [Brief explanation of the drawings]

[0030] [Figure 1] FIG. 1 is a block diagram showing an example of a robot control system according to an embodiment of the present invention. [Figure 2]FIG. 2 is an illustrative view showing an example of the communication robot of the embodiment of FIG. [Figure 3] FIG. 3 is a block diagram showing an example of the electrical configuration of the communication robot of the embodiment shown in FIG. [Figure 4] FIG. 4 is a block diagram showing the electrical configuration of the operation terminal in the embodiment shown in FIG. [Figure 5] FIG. 5 is an illustrative view illustrating a bowing action for apologizing in an embodiment. [Figure 6] FIG. 6 is a diagram illustrating the results of an empirical linear regression of the relationship between the severity of the failure and the forward tilt angle (θb) of the bowing motion. [Figure 7] FIG. 7 is a diagram showing the results of an empirical linear regression of the relationship between the severity of the failure and the forward lean speed (tb) during bowing. [Figure 8] FIG. 8 is a diagram showing the results of empirical linear regression on the relationship between the severity of failure and the duration of forward lean (tk) in bowing movements. [Figure 9] 9 is an illustrative view showing one example of a memory map of a RAM of the memory of the operation terminal shown in FIG. 4. FIG. [Figure 10] 9 is an illustrative view showing an example of a memory map of the RAM of the memory of the robot shown in FIG. 3. FIG. [Figure 11] FIG. 11 is a flow chart showing an example of the operation of the operation terminal in the embodiment of FIG. [Figure 12] FIG. 12 is a flow chart showing an example of generating the robot's motion in the embodiment of FIG. [Figure 13] FIG. 13 is a flow chart showing an example of the bowing action performed by the robot in the embodiment of FIG. [Figure 14] FIG. 4 is an illustrative view showing one example of a memory map of the RAM of the memory of the robot shown in FIG. 3 in the second embodiment. [Figure 15] FIG. 15 is a flowchart showing an example of the operation of the robot in the second embodiment. [Figure 16] FIG. 16 is a flowchart showing an example of motion generation for a robot in the second embodiment. BEST MODE FOR CARRYING OUT THE INVENTION

[0031] 1, a robot control system 10 of this embodiment includes a social robot or a communication robot (hereinafter, sometimes simply referred to as a "robot") 12. This robot 12 is a robot that can communicate with a user face-to-face.

[0032] In this embodiment, the robot 12 is a humanoid robot as shown in FIG. 2. However, it should be noted that any robot with a different appearance and structure can be used as the robot 12 used in the embodiment of FIG. 1. However, the robot must be able to lean its upper body forward, for example, so that it can perform an apology action by bowing. Even if the upper body does not have a structure that allows it to lean forward, a robot that can lean its head forward can also be used.

[0033] The robot 12 is connected to an operation terminal 16, which serves as a remote control device, via a network 14 such as the Internet or a telephone communication line. The operation terminal 16 is a general-purpose computer such as a PC, PDA, smartphone, or tablet terminal. The operation terminal 16 may be installed near the robot 12.

[0034] 1, the operation terminal 16 is further connected to a microphone 18 that collects the operator's voice when the operator interacts with a user through the robot 12, and a camera 20 that captures images of the environment and, in some cases, images of the user. Image data from the camera 20 is input to the operation terminal 16, and if necessary, the operation terminal 16 can process the image data to, for example, identify the user.

[0035] 1, the network 14 can be connected to a cloud server 22. In this embodiment, the cloud server 22 configures a large-scale language model, such as ChatGPT (product name). The operation terminal 16 can input prompts (sentences indicating questions or instructions) into the large-scale language model of the cloud server 22 when necessary and use the model.

[0036] Referring to FIG. 2, the social robot or communication robot 12 in this embodiment is, as an example, a humanoid robot called "PEPPER" (registered trademark) provided by SoftBank Robotics Corp. However, it should be noted in advance that the present invention can be applied not only to three-dimensional robots such as humanoid robots, but also to two-dimensional robots (CG agents) created using CG (computer graphics). Therefore, the term "robot" should be understood to encompass not only robots but also CG agents. However, all robots must be capable of bowing, i.e., their upper body and / or head must be able to lean forward.

[0037] The robot 12 shown in FIG. 2 includes a movable base 24, on which are mounted legs 26 corresponding to the legs or feet of a human being, and on which are mounted, via waist or hip joints 30, a torso 28 constituting part of the upper body.

[0038] The torso 28 is a part corresponding to the torso of a human being, and left and right arms 32 are attached to both shoulders of the torso 28 via shoulder joints, and hands 33 are attached to the ends of the arms 32.

[0039] A display 34 is provided on the front of the torso 28, and a head 38 is provided above the torso 28 via a neck 36. The display 34 can display necessary images when it is desired to communicate information. However, the display 34 may be a touch display equipped with a touch panel.

[0040] The head 38 has a built-in microphone 54 (FIG. 2) in the area corresponding to the mouth, and a built-in stereo camera 50 (FIG. 2) in the area corresponding to the eyes.

[0041] In addition, although not shown, the robot 12 is equipped with various sensors for collision prevention and distance measurement, as well as contact sensors, but since these sensors are not relevant to this embodiment, detailed explanations thereof will be omitted.

[0042] As shown in Fig. 3, the robot 12 includes a CPU 40 that controls the overall operation of the robot 12. The CPU 40 is connected to a memory 42 and a communication unit 44 via a bus 41. The memory 42 includes, for example, a RAM and an HDD (hard disk drive). In this embodiment, the communication unit 44 can perform necessary communication with an external device, such as the operation terminal 16, according to, for example, Wi-Fi (as an example, IEEE 802.11 a / b / g / n).

[0043] In this embodiment, all operations of the robot 12 are performed in response to commands (including control data) from the CPU 40. That is, the CPU 40 provides commands to the actuator control circuit 45 via the bus 41, and the actuator control circuit 45 controls the operation of each of the actuators A1-An in accordance with the command values ​​corresponding to the commands. For example, the actuator control circuit 45 generates pulse power in the number corresponding to the command values ​​provided by the CPU 40 and provides the pulse power to the corresponding stepping motor to drive each of the actuators A1-An.

[0044] However, in addition to the actuator using such a stepping motor, any actuator such as an actuator using a servo motor or a fluid actuator can be used.

[0045] Here, one of the actuators A1-An is an actuator (digital servo) of the waist joint or hip joint 30 shown in Fig. 1, and the others are actuators of other joints. Such actuator control circuit 45 controls actuators A1-An in accordance with commands from CPU 40, thereby enabling robot 12 to perform a bowing motion (movement of leaning the upper body forward) for apologizing.

[0046] The sensor I / F (interface) 46 is connected to the CPU 40 via the bus 41, and receives outputs from the various sensors 48 and camera 50 described above.

[0047] The camera 50 constitutes part of the vision of the robot 12. In other words, the camera 50 is used to detect video or images seen by the eyes of the robot 12. In this embodiment, data (image data) corresponding to the video (moving or still image) captured by the camera 50 is provided to the CPU 40 via the sensor I / F 46, and further, if necessary, is sent from the communication unit 44 to the operation terminal 16 via the network 14.

[0048] Furthermore, the speaker 52 and microphone 54 are connected to the input / output I / F 51. The speaker 52 outputs voice when the operator of the operation terminal 16 speaks through the robot 12, for example.

[0049] The microphone 54 constitutes part of the hearing of the robot 12. This microphone 54 is mainly used to detect the voice of a user (person) who is interacting (communicating) with the robot 12. That is, in this embodiment, an audio signal corresponding to the voice from the microphone 54 is given to the CPU 40 via the input / output I / F 51, and further, if necessary, is sent from the communication unit 44 to the operation terminal 16 via the network 14.

[0050] 4, the operation terminal 16 in the embodiment of FIG. 1 includes a CPU 56 that is responsible for overall control of the operation terminal 16, and a memory 58 is connected to the CPU 56 via a bus 57. The memory 58 includes RAM, ROM, HDD, etc. The CPU 56 can remotely control the robot 12 by executing a computer program described below. The program is stored in advance in the ROM or HDD, and is loaded into the RAM and executed as needed.

[0051] An input interface 60 is further connected to the bus 57. The microphone 18 shown in Fig. 1 is connected to the input interface 60, and for example, an operator's voice signal from the microphone 18 is input from the input interface 60 to the CPU 56 (memory 58).

[0052] The input interface 60 is further connected to the camera 20, and the video signal from the camera 20 is input from the input interface 60 to the CPU 56 (memory 58).

[0053] A communication unit 62 is further connected to the bus 57, and this communication unit 62 communicates with the robot 12, the cloud server 22, etc. via the network 14. For example, it is used to receive data transmitted from the robot 12 (video signals from the camera 50, audio signals from the microphone 54, and electrical signal data from the various sensors 48) and to transmit various commands to the robot 12.

[0054] In this embodiment, an audio signal from the microphone 20 is input to the CPU 56 through the input interface 60, and the audio signal is further sent from the communication unit 62 via the network 14 to the robot 12 (communication unit 44 thereof).

[0055] An input device 62 such as a keyboard or a touch panel is also connected to the input interface 60. Using this input device 62, the operator can create sentences expressing examples of failures by the robot 12, which will be described later, and input these sentences into a large-scale language model such as ChatGPT on the cloud server 22 (FIG. 1).

[0056] An output interface 64 is also connected to the bus 57. The output interface 64 outputs, for example, a video signal from the camera 20 or a video signal from the robot 12 to a display 66 to display a video required for the operator. The output interface 64 outputs, for example, an audio signal from the microphone 18 or an audio signal from the robot 12 to a speaker 68 to output a sound required for the operator.

[0057] When an operator remotely controls the robot 12 using the operation terminal 16, the robot interface 70 transmits commands (instructions including setting data or control data) from the CPU 56 to the CPU 40 (Figure 3) of the robot 12 to control the movement of the robot 12.

[0058] As explained above, when a robot makes a mistake, no specific guidelines have been presented for designing an appropriate bowing motion that corresponds to the severity of the mistake when a robot apologizes.

[0059] Therefore, in this embodiment, a robot and its control system are provided that can perform a bowing motion to appropriately express an apology depending on the severity of the robot's failure, which will affect its social acceptance.

[0060] For example, a person may bow lightly for a less serious mistake, such as being a few minutes late, whereas a more serious mistake, such as injuring someone else, may bow deeply and maintain that position for several seconds.

[0061] Based on these considerations, in order to develop a model for calculating appropriate bowing movements for a robot to express an apology, the inventors first collected data on the degree, speed, and length of bowing to appropriately express an apology based on various levels of seriousness.

[0062] Using the collected data, the inventors developed a polynomial regression model to identify appropriate bowing behaviors based on the severity of the failure, and experimentally investigated the effectiveness of the developed model in terms of tolerance and reliability.

[0063] For example, since it was found difficult to directly compare various failures such as lateness and financial loss, the inventors considered the severity of the apology situation in a unified manner. Ultimately, we decided to focus on three different apology situations: time, money, and injury, recognizing that losses in these areas typically prompt individuals to apologize to those harmed.

[0064] Therefore, we designed five different cases for each situation, taking into account the various levels of severity within these categories, and devised one questionnaire item to evaluate the severity of each apology scenario.The questionnaire consisted of 11 response items.

[0065] For example, five cases were prepared for the time loss when waiting for a meeting: 5 minutes, 15 minutes, 30 minutes, 60 minutes, and more than 60 minutes. These five items were heuristically determined through multiple discussions so that the values ​​would be relatively equal among the inventors when the severity was divided into 11 levels.

[0066] For example, for monetary losses, we identified five scenarios ranging from a few dollars to tens of thousands of dollars: a few dollars, tens of dollars, hundreds of dollars, thousands of dollars, and tens of thousands of dollars. Again, this category was decided upon heuristically after extensive discussion, aiming for a fairly uniform distribution across the defined range.

[0067] For example, for injuries inflicted on people, we outlined five levels of injury severity: minor injury not requiring hospitalization, moderate injury requiring hospitalization without life-threatening complications, severe injury requiring hospitalization but not immediately life-threatening, severe with imminent life-threatening consequences, and fatal outcome (death) (expected apologies to the family). These categories were also established heuristically after thorough discussion.

[0068] In the actual data collection, the inventors collected data on the timing and angle of the bowing motion. As shown in Figure 5, they focused on four important variables: tb, tk, te, and θb. These variables represent the time tb required to bow (forward lean speed tb), the time tk required to maintain the bow (forward lean duration tk), the time te required to return to the original position after completion (recovery speed te), and the angle θb (forward lean angle θb). However, the tb and te values ​​are equal, indicating that the time required to bow and the time required to return to the original position after completion of the bowing motion are the same.

[0069] For example, to explain the bowing action shown in Figure 5 in relation to the robot 12 shown in Figure 2, when the robot 12 realizes that it (the robot) has failed, for example, in response to an instruction from the operation terminal 16 or of its own judgment, the CPU 40 issues a command to the actuator control circuit 45 and controls the waist joint or hip joint 30 by the corresponding actuator (Figure 3) so that the torso 28 is at a forward lean angle θb relative to the legs 26 before the forward lean time tb has elapsed.

[0070] In this embodiment, the neck 36 does not tilt the head 38 forward, but bowing movements that tilt the head 38 forward are not excluded.

[0071] After the angle of the hip joint 30 reaches a state corresponding to the forward tilt angle θb, the CPU 40 issues a command to the actuator control circuit 45, and the actuator controls the hip joint 30 so that the angle is maintained for the forward tilt duration tk.

[0072] After the forward tilt duration tk has elapsed, the CPU 40 issues a command to the actuator control circuit 45, and the actuator controls the hip joint 30 so that the hip joint 30 returns to its original position during the recovery time te.

[0073] The data collection carried out by the inventors involved 20 participants, 10 males and 10 females, all of whom were native Japanese speakers. The mean age was 21.05 years, with a standard deviation of 1.91.

[0074] Data collection consisted of participants' perceived seriousness and bowing behavior. As mentioned above, the former was measured using a single questionnaire item that assessed the seriousness of the apology situation using 11 answers. For the latter, three parameters were measured: the bow angle θb, the time until bowing tb, and the duration tk. To do this, a depth camera (for example, Intel's Real Sense Depth Camera, product name "D435") and Nuitrack, a skeleton tracking software development kit (SDK), were used to calculate the position of each joint.

[0075] Each participant was given a brief explanation of the purpose and procedure of the experiment beforehand, and written informed consent was obtained from all participants.

[0076] When data collection began, participants were instructed to imagine themselves in an apology scenario. They then bowed and verbally apologized (e.g., "I'm sorry") at the beginning of the bowing motion. These bowing motions were captured using a depth camera. Participants randomly performed five cases for each scenario, balancing the order of the apology scenarios and the cases in each scenario. After each bow, participants answered a questionnaire and repeated the bowing motion 15 times.

[0077] Next, we will explain the results of the data analysis on bowing behavior in the apology scenario.

[0078] First, we analyzed the severity questionnaire items (shown in Table 1). Note that the numbers in parentheses in Table 1 are standard deviations.

[0079] [Table 1]

[0080] With the exception of the injury scenario, the severity ratings of the five scenarios across the three situations showed a varied distribution. Furthermore, the average severity rating of the mildest case (5 minutes late) was approximately 2.25. Although certain cases showed similar severity values, especially higher ones for the injury scenario, the majority of cases were evenly distributed along the severity axis. Therefore, all cases were analyzed uniformly along that axis. Next, we explain the relationship between the severity of bowing movements in all situations, the degree θb, the speed or time tb and te, and the duration tk.

[0081] We hypothesized that there is a positive correlation between severity and bow angle θb, and conducted a polynomial regression analysis. The results are shown in Figure 6. In this analysis, the intercept was set to 20 degrees, in line with previous research suggesting that the bow range is 20-30 degrees. This regression analysis shows how the bow angle, i.e., the forward tilt angle θb, changes depending on severity. We derived the following equation (1) for the forward tilt angle θb.

[0082]

number

[0083] Next, we performed a polynomial regression analysis, assuming a positive correlation between the severity and the time to bow (forward lean speed) tb, as well as the bow angle. The results are shown in Figure 7. The intercept was set to 0.5 seconds to reflect the time required for the robot's 12 actuators to control the motors, and 0.5 seconds was used as the minimum value for modeling. The forward lean speed is affected by the bow angle. This is because a larger angle requires more time to complete the bowing motion than a smaller angle. We derived the following equation (2) for the forward lean time tb:

[0084]

number

[0085] Finally, assuming a positive correlation between severity and duration tk, we performed a polynomial regression analysis. The results are shown in Figure 8. In this analysis, the intercept was set to 0 (zero) seconds, based on the assumption that bowing may not be maintained in less serious situations. The results of the polynomial regression analysis show how bowing duration is adjusted depending on severity. We derived the following equation (3) for bowing duration tk:

[0086]

number

[0087] Thus, it became clear that the severity of the apology situation significantly affected the participants' bowing behavior, particularly at the forward tilt angle θb, forward tilt time tb, and forward tilt duration tk points. Our analysis demonstrated that a polynomial regression approach effectively models these characteristics. Therefore, in the example described below, the robot's apologetic bowing behavior is controlled according to these models.

[0088] Fig. 9 is a diagram showing an example of a memory map of the memory 58 of the operation terminal 16 shown in Fig. 4. The memory 58 includes a RAM as described above, which includes a program storage area 72 and a data storage area 74 as shown in Fig. 9. The program storage area 72 stores a control program for the operation terminal 16. The control program and necessary data may be stored in advance in, for example, a flash memory or a hard disk included in the memory 58, and may be read out as needed and expanded into the RAM.

[0089] The control programs of the operation terminal 16 include a display control program 76a, an operation detection program 76b, a voice recognition program 76c, a failure determination program 76d, a severity determination program 76e, and the like.

[0090] The display control program 76a is a program that causes the display 66 (FIG. 4) to display an image.

[0091] The operation detection program 76b is a program that detects an operation on the input device 62 of the operation terminal 16 (FIG. 4).

[0092] The voice recognition program 76c is a program for voice recognition of a voice signal, for example, when the voice signal is collected by the microphone 54 (FIG. 3) of the robot 12 and transmitted through the communication units 44 and 62. The voice data resulting from the voice recognition is converted into character data and stored, for example, in a temporary storage area 78c (described later) of the data storage area 74.

[0093] However, it is also possible to use commercially available software such as Google (trademark) speech recognition as the speech recognition program 76c.

[0094] The failure determination program 76d is a program for determining whether the robot 12 has failed, and determines that the robot 12 has failed by judging the user's speech picked up by the microphone 54 (FIG. 3) of the robot 12, for example, a voice such as "You're late. You're 10 minutes late." This voice "You're late. You're 10 minutes late." reminds the robot 12 that it is late, and in terms of the categories mentioned above, this is a time failure.

[0095] Other possible voices could be, for example, a voice evoking a financial failure such as, "I dropped a 1000 yen glass cup and broke it..." or a voice evoking a failure related to injury such as, "I was injured by the fence that the robot hit and knocked over. I'm bleeding a little."

[0096] The severity determination program 76e is a program for determining the severity of the failure determined by the failure determination program 76d. Specifically, in accordance with this severity determination program 76e, the operation terminal 16 uses the input device 62 to input failure cases into a large-scale language model on the cloud server 22 as text data that describes the cases, and has the large-scale language model such as ChatGPT determine (determine) the severity of the failure.

[0097] For example, "Please express the severity of failure in the following situations using a number between 0 and 10. As examples of standards, the severity of being 5 minutes late would be 2.25, the severity of causing a loss of $10 would be 5.25, and the severity of causing the other person's death would be 9.95. Sentences such as the failure situation "I dropped a product worth $20 and made the other person wait for more than 10 minutes" are input into a large-scale language model as text data.

[0098] Another example is when someone damages personal belongings or pays compensation, and the severity is estimated using a large-scale language model based on the amount. An example prompt for this is, "Please express the severity of failure in the following situations using a number between 0 and 10. For example, the severity of a loss of a few dollars is 3.3, the severity of a loss of tens of dollars is 5.25, the severity of a loss of hundreds of dollars is 7.7, the severity of a loss of thousands of dollars is 8.85, and the severity of a loss of tens of thousands of dollars is 9.65." A possible failure situation might be something like, "I damaged $350 worth of personal belongings."

[0099] However, the first half of the document, which mentions examples of the severity criteria, can be a standard phrase, and the second half, which describes the details of the failure, can be changed depending on the failure case.

[0100] Regarding physical injuries, instead of querying the large-scale language model of the cloud server 22, it is also possible to have the user provide an electronic medical record and determine the severity from the information therein. Five levels are defined as follows: for example, minor injury (no need for hospitalization, 5.55), moderate injury (not life-threatening but requiring hospitalization, 8.50), severe injury (potentially life-threatening, 9.55), critical injury (imminent life-threatening, 9.75), and death (requiring apology to the victim's family, 9.95).

[0101] In the data storage area 74, an image generation data storage area 78a, an operation data storage area 78b, a temporary storage area 78c, and the like are formed.

[0102] Image generation data storage area 78a stores image generation data including polygon data, texture data, and other data that are set in advance to generate data for various screens to be displayed on display 66.

[0103] The operation data storage area 78b stores data input from the input device 62 in chronological order.

[0104] The temporary storage area 78c is an area for temporarily storing data such as text data (character data) representing failure cases to be input into the large-scale language model of the cloud server 22 and severity data returned from the large-scale language model.

[0105] The data storage area 74 stores other data required for the execution of the control program of the operation terminal 16, and also includes flags and other counters (timers) required for the execution of the control program.

[0106] Fig. 10 is a diagram illustrating an example of a memory map of the memory 42 of the robot 12 shown in Fig. 3. The memory 42 includes a RAM as described above, which includes a program storage area 80 and a data storage area 82 as shown in Fig. 10. The program storage area 80 stores a control program for the robot 12. The control program and necessary data may be stored in advance in, for example, a flash memory or a hard disk included in the memory 42, and may be read out as needed and loaded into the RAM.

[0107] The control program for the robot 12 includes a display control program 84a, an operation detection program 84b, a voice recognition program 84c, an apology program 84d, and the like.

[0108] The display control program 84a is a program that displays an image on the display 34 (FIG. 2).

[0109] The operation detection program 84b is a program that detects an operation on the input device of the robot 12 (for example, the touch panel of the display 34).

[0110] The voice recognition program 84c is a program that converts voice data obtained by recognizing voice from the microphone 54 into character data, and the voice data is stored in a temporary storage area 86c (described later) of the data storage area 82 as necessary.

[0111] However, it is also possible to use commercially available software such as Google (trademark) speech recognition as the speech recognition program 84c.

[0112] The apology program 84d is a program for making an apologetic bow to a user or the like when the robot 12 makes a mistake, depending on the severity of the mistake (for example, as shown in Table 1), and includes, for example, a motion generation program 84d1 shown in FIG. 12 and an execution program 84d2 shown in FIG. 13 for executing the bowing motion generated by the motion generation program 84d1.

[0113] In the data storage area 82, an image generation data storage area 86a, an operation data storage area 86b, a temporary storage area 86c, and the like are formed.

[0114] The image generation data storage area 86a stores image generation data including polygon data, texture data, and other data that are set in advance to generate data for various screens to be displayed on the display 34.

[0115] The operation data storage area 86b stores data input from an input device (not shown) in chronological order.

[0116] The temporary storage area 86c is an area for temporarily storing the results of voice recognition by the voice recognition program 84c, the severity data sent from the operation terminal 16, and the like, as described above.

[0117] The data storage area 74 stores other data required for executing the control program for the robot 12, and also includes flags and other counters (timers) required for executing the control program.

[0118] In the first embodiment, the operation terminal 16 operates according to the flow chart shown in FIG.

[0119] In the first step S1 of Fig. 11, the CPU 56 (Fig. 4) determines whether or not the robot 12 has failed. That is, when the robot 12 has failed, an operation on this operation terminal 16 is started.

[0120] As described above, whether the robot 12 has failed is determined based on the result of voice recognition of the user's utterance picked up by the microphone 54 of the robot 2 (FIG. 3).

[0121] When the CPU 56 determines that the robot 12 has failed as a result of the voice recognition, in the following step S2, in accordance with the severity determination program 76e, as described above, it sends text data that documents the failure case to the cloud server 22 (Figure 1) via the communication unit 44 and the network 14, and asks the cloud server 22 to determine the severity of the failure caused by the robot 12.

[0122] Thereafter, when the severity data is sent from the cloud server 22 in step S3, the CPU 56 stores the severity data in the temporary memory area 78c and transmits the severity data to the CPU 40 of the robot 12 via the communication unit 62, the network 14 and the communication unit 44.

[0123] Thereafter, in step S5, the CPU 56 of the operation terminal 16 determines whether the apology accompanied by the bowing action of the robot 12 has been accepted by the relevant user (the person who received the mistake).

[0124] The determination of whether the apology in step S5 has been accepted is made, for example, by the operator of the operation terminal 16 based on the video and / or audio transmitted from the robot 12. For example, the operator remotely determines on the operation terminal 16 whether the other person's angry face has turned into a smile, or whether the other person has accepted the apology or made a positive utterance.

[0125] The action generation program 84d1 shown in FIG. 12 according to the apology program 84d of the robot 12 starts when the severity data is notified from the operation terminal 16 in step S4.

[0126] In the first step S11 of Fig. 12, the CPU 40 of the robot 12 calculates the forward lean angle θb based on the severity. In this step S11, the forward lean angle θb is calculated according to the above-mentioned formula (1). For example, if the robot 12's failure is 30 minutes late and the severity is 5.60 (Table 1), the forward lean angle θb can be calculated as θb = -0.447 × 5.60 × 5.60 + 10.358 × 5.60 + 20 ≒ 64.0 degrees.

[0127] In the next step S12, the CPU 40 of the robot 12 calculates the forward lean time (forward lean time) tb based on the severity. In this step S12, the forward lean time tb is calculated according to the above-mentioned formula (2). If the severity is assumed to be 5.60, the forward lean time tb can be calculated as tb = -0.007 × 5.60 × 5.60 + 0.1446 × 5.60 + 0.5 ≒ 1.09 seconds.

[0128] In the next step S13, the CPU 40 of the robot 12 calculates the forward lean duration tk based on the severity. In this step S12, the forward lean duration tb is calculated according to the above-mentioned formula (2). If the severity is assumed to be 5.60, the forward lean duration tk can be calculated as tk = 0.0216 × 5.60 × 5.60 - 0.0472 × 5.60 ≒ 0.41 seconds.

[0129] Thereafter, the CPU 40 stores these numerical values ​​in the temporary storage area 86c (FIG. 10) and ends this motion generation subroutine.

[0130] Next, the robot 12 executes the execution program 84d2 shown in FIG.

[0131] In the first step S21 of Fig. 13, the CPU 40 determines whether or not the bowing motion can be executed as is. Since the bowing motion for apologizing is based on the bowing motion of a human, in the case of a robot such as the robot 12 in the embodiment, which has a range of motion different from that of a human, motion conversion is necessary.

[0132] The maximum range of motion of the hip joint 30 of PEPPER (product name), which is the robot 12 used in the embodiment, is 59.5 degrees. On the other hand, in the example of Table 1, the maximum forward tilt angle θb calculated from the calculation of equation (1) is 78.8 degrees. In other words, the robot 12 of the embodiment cannot achieve the calculated maximum forward tilt angle. Therefore, in step S21, the CPU 40 determines whether the forward tilt angle θb indicated by the calculation result is greater than or smaller than the maximum forward tilt angle (maximum range of motion) of the robot 12, thereby determining whether the bowing motion can be performed as is.

[0133] The forward tilt angle θb≈65.0 degrees exemplified above is greater than the maximum forward tilt angle (maximum range of motion) of the robot 12 described above, and therefore "NO" is determined in step S21.

[0134] When the calculated forward tilt angle θb is greater than the maximum forward tilt angle (maximum range of motion) of the robot 12, the robot 12 cannot be made to perform a bowing motion at that calculated forward tilt angle.

[0135] Therefore, in this embodiment, in step S22, the CPU 40 calculates a forward tilt angle suited to the robot body, where the robot body refers to a robot that apologizes by bowing, etc.

[0136] When a human performs a bowing motion, the maximum forward tilt angle Dhmax obtained by the calculation of equation (1) is set to 78.8 degrees, and the minimum forward tilt angle Dhmin is set to 20.0 degrees. Furthermore, the maximum forward tilt angle Drmax of the robot 12 in this embodiment is set to 59.5 degrees, and the normalization coefficient Rscale is calculated using the following equation (4). The forward tilt angle θb is converted to the robot forward tilt angle θrb using the normalization coefficient Rscale using the following equation (5).

[0137] [Number 4] Rscale=(Drmax-Dhmin) / (Dhmax-Dhmin) (4)

[0138]

number

[0139] As illustrated above, when θb is calculated to be 65.0 degrees from equation (1), the normalization coefficient Rscale is approximately 0.67177, and the robot forward tilt angle θrb can be calculated to be approximately 30.23 degrees.

[0140] Therefore, in the example shown, in the next step S23, the forward lean angle θb is changed to the robot forward lean angle θrb, and the bowing action is performed with the forward lean time tb≒1.09 seconds, forward lean angle≒30.23 degrees, forward lean duration tk≒0.41 seconds, and recovery time te≒1.09 seconds calculated in FIG. 12.

[0141] However, if the subject apologizing is a two-dimensional agent created using CG, there is no limit to the forward tilt angle θb, so the determination in step S21 is "YES."

[0142] If the answer to step S21 is "YES," then in step S23, a series of bowing movements is executed, in which the user leans forward at a forward lean angle θb for a forward lean time tb calculated in FIG. 12, maintains this state for a duration tk, and returns to the original position for a return time te.

[0143] Although not shown in the figures, when performing the bowing action for apologizing according to the flow chart of Figure 13, it is desirable to utter an apology, such as "I'm very sorry," before or along with the bowing action.

[0144] In this way, in the above-described embodiment, the robot 12 calculates the parameters tb, θb, tk, and te using equations (1)-(3) based on the designated severity, and controls the bowing motion of the robot 12 according to these parameters, but if necessary, modifies the parameter θb using equations (4)-(5) to generate the bowing motion. In the above-described embodiment, the severity of the failure is provided externally.

[0145] Prior to the experiment, the inventors collected data to investigate the relationship between bowing parameters and seriousness, and analyzed the data. The results showed that as seriousness increased, bowing became deeper and longer, and they gained the knowledge that polynomial regression effectively captured changes in parameters based on seriousness.

[0146] We hypothesized that bowing behaviors that match the level of seriousness would be perceived as more acceptable than bowing behaviors that ignore the level of seriousness. Based on this hypothesis, we made the following predictions. Prediction 1: When a robot apologizes for a mistake with a bow that takes into account the seriousness of the mistake, people are more likely to accept the apology and form a more positive impression of it than when the bow ignores the seriousness of the mistake. Prediction 2: When a robot's apology for a mistake includes a bowing gesture that takes into account the seriousness of the mistake, trust in the robot will increase compared to when the bowing gesture ignores the seriousness of the mistake.

[0147] Based on this prediction, the inventors created a video in which a robot bows and apologizes for its mistake. One element (a modeling element consisting of a proposed condition and an alternative condition) was set up, and scenarios such as 5-minute lateness, 15-minute lateness, minor injury, financial loss of several hundred dollars, and serious injury were represented in a total of 10 videos, five for each condition. These five videos were selected because they vary in severity, with ratings of 2.25, 3.75, 5.55, 7.70, and 9.75, respectively.

[0148] In the proposed condition, the bowing motion of the robot 12 is performed according to the parameters calculated by the above formula, and in the alternative condition, the forward lean angle θrb is 45 degrees, the forward lean time tb is 1.5 seconds, and the forward lean duration tk is 2 seconds. Therefore, in the alternative condition, the robot performs the same bowing motion regardless of the change in severity.

[0149] Using these videos, 200 participants were surveyed to investigate the effectiveness of the robot's apology. The results confirmed the effectiveness of the bowing gesture, in line with the two predictions above.

[0150] In the above-described embodiment, the operator of the operation terminal 16 assigns a seriousness level to the robot 12, and the robot 12 calculates the parameters of an appropriate bowing action when apologizing based on the assigned seriousness level. However, since this function is clearly important for realizing the autonomous generation of an apologetic behavior, a second embodiment in which the robot 12 autonomously estimates the seriousness level will be described below.

[0151] In the second embodiment, a failure determination program 84e is incorporated into the memory 42 of the robot 12 as shown in Fig. 14, and the motion generation program 84d1 of the apology program 84d is modified as shown in Fig. 16. Furthermore, a table data storage area 86d is provided in the data storage area 82 of the memory 42, and the data in Table 1 is stored therein. This is because it is necessary for the robot 12 to autonomously estimate the severity.

[0152] In the second embodiment, the CPU 40 of the robot 12 determines whether the robot 12 has failed in step S1 shown in Fig. 15. This determination method may be the same as the determination performed by the operation terminal 16 described above with reference to Fig. 11. For example, whether the robot 12 has failed is determined based on the result of voice recognition of the user's utterance picked up by the microphone 54 (Fig. 3) of the robot 12.

[0153] Alternatively, as in the previous embodiment, the operation terminal 16 may determine whether or not the operation has failed, and when the operation terminal 16 determines that the operation has failed, it may notify the robot 12 of this.

[0154] Thereafter, the CPU 40 of the robot 12 executes a subroutine shown in FIG. 16 in step S6 of FIG.

[0155] In the first step S01 of Fig. 16, the CPU 40 performs image information processing and / or audio information processing using, for example, deep learning, to recognize the cost of the failure, such as the time loss, the amount of money loss, or the degree of injury in terms of physical damage.

[0156] As another example, estimating the delay of a robot 12 can be easily achieved by integrating a scheduling system with a timer. For example, if a robot 12 is late for a meeting with a human or misses a delivery pickup time, the severity of the delay can be easily estimated.

[0157] Regarding financial damage, image processing-based object recognition systems, such as the Yolo Algorism, can help estimate the financial loss of failure. For example, these image processing systems can effectively identify high-value items such as jewelry and computers.

[0158] Similarly, vision-based approaches have the potential to assess damage, for example, the recent promise of multimodal large-scale language models to support diagnosis with clinical imaging data.

[0159] Then, in step S02, the CPU 40 estimates the severity of the cost. For example, the CPU 40 uses the relationship between cost and severity shown in Table 1, which is stored in advance in area 86d of the data storage area 74 of the memory 42 of the robot 12. In this case, the costs of financial, time, and physical damage are compared with those in Table 1, and the approximated value is used as the severity.

[0160] Another thing to consider is estimating the intensity of the person you are talking to. Even if the severity is minor, if the other person's anger increases, it may be better to apologize more seriously.

[0161] 15, the robot 12 autonomously determines a failure and determines the severity in step S7 (steps S01 and S02). However, in this second embodiment, as in the previous embodiment, it is also possible for the robot 12 to have the large-scale language model of the cloud server 22 determine the severity. In this case, it is conceivable that the robot 12 transmits voice data to the cloud server 22.

[0162] Thereafter, similarly to the previous embodiment, the CPU 40 calculates the forward tilt time tb, the forward tilt angle θb, and the forward tilt duration tk in steps S11, S12, and S13, respectively, and executes the subroutine shown in FIG. 13 in the next step S7 using the calculation results as return values.

[0163] The subroutine of step S8 has been explained above with reference to FIG. 13, so a duplicate explanation will be omitted here.

[0164] In the final step S9 of Fig. 15, it is determined whether the apology has been accepted. The determination method can be the same as that used by the operation terminal 16 in step S5 of Fig. 11 in the previous embodiment. For example, the determination can be made based on the video and / or audio of the user (other party) input from the camera 50. For example, the CPU 40 determines whether the other party's angry face has changed to a smile, or whether the other party has accepted the apology or made a positive statement.

[0165] However, if the determination in step S9 is "NO," the CPU 40 increases the severity level by one level in the same failure category in step S10, and executes step S5 and subsequent steps again. For example, in the above example, if the severity level initially determined in step S5 for a time-related failure was "5.60" (30 minutes late), but the determination in step S9 is "NO," the CPU 40 sets the severity level to "7.75," one level higher, in the same category (time), and returns to step S5.

[0166] This can also be applied to the first embodiment described above. As shown in Fig. 11, when the operation terminal 16 determines "Was the apology accepted?" in step S5 and the CPU 56 determines "NO," the seriousness level within the same failure category can be increased by one level, as in step S10 in Fig. 15, and the updated seriousness level can be transmitted again to the robot 12 in step S4 in Fig. 11. At this time, the robot 12 can re-execute the behavior generation routine shown in Fig. 12 in accordance with the updated seriousness level.

[0167] In this way, even if the apology is not accepted, by increasing the seriousness and performing the bowing action again according to the parameter value corresponding to that seriousness, the apology is more likely to be accepted in the end.

[0168] In all of the above-described embodiments, the bowing motion of the robot 12 is controlled using the forward lean time (forward lean speed) tb, forward lean angle θb, forward lean duration tk, and recovery time (recovery speed) te calculated in steps S11-S13 of FIG. 12 or FIG. 16, but it is not necessary to control the bowing motion according to all of the parameters.

[0169] For example, since the depth of the bow best expresses the seriousness, the bowing action of the robot 12 can be controlled using the forward lean angle calculated based on the seriousness.

[0170] Furthermore, the duration of the bow, i.e., the duration of the forward lean, is an easy way to express the seriousness, so the bowing action of the robot 12 can be controlled using the duration of the forward lean calculated based on the seriousness.

[0171] In both of the above-mentioned two embodiments, the robot 12 capable of forward leaning motion of the upper body is used to perform a bowing motion. In this case, when the maximum forward leaning angle of the robot 12 is smaller than the forward leaning angle calculated based on the severity, the forward leaning angle converted by a conversion formula such as formula (5) is set.

[0172] For example, this invention can be applied to an android robot that resembles a human, but if there is a similar restriction on the forward lean angle, it is necessary to set a similarly converted forward lean angle.

[0173] However, if a humanoid CG agent is used as the robot 12, there is no limit to the forward lean angle, so it is possible to directly execute a bowing action with a forward lean angle calculated based on the severity, and therefore there is no need to perform conversion calculations for the forward lean angle to match the robot body.

[0174] Furthermore, when using, for example, CommU (product name) or Sota (product name) manufactured by Vstone Corporation as the robot 12, these robots do not have a movable axis in the waist, so the forward tilt angle described in the embodiment is expressed as the forward tilt angle of the head. The forward tilt angle can be calculated using the above formulas (4) and (5) by setting the minimum head angle to 20 degrees (Drmin) and the maximum head angle (Drmax) to 30 degrees.

[0175] The processing of each step in the flow diagrams shown in FIGS. 11 to 13 and 15 to 16 is merely an example, and the processing order of each step may be changed as long as the same results are obtained.

[0176] Furthermore, the specific values ​​of the angles, durations, etc. given above are merely examples and can be changed as needed. [Explanation of symbols]

[0177] 10...Robot control system 12...Communication robot 30...Hip joint 40, 56...CPU 42, 58...Memory 45... Actuator control circuit

Claims

1. A robot control system for controlling a robot capable of bowing by tilting at least one of its upper body and head forward, comprising: a severity determination unit that, when the robot fails, determines the severity of the failure in a category; a parameter value calculation unit that calculates a parameter value for controlling the bowing action based on the severity determined by the severity determination unit; and A robot control system comprising a motion control unit that controls the bowing motion of the robot in accordance with the forward lean parameter value calculated by the parameter value calculation unit.

2. the parameter value calculation unit includes a forward lean angle calculation unit that calculates a forward lean angle of the robot based on the severity; The robot control system according to claim 1 , wherein the movement control unit controls the bowing movement of the robot in accordance with the forward tilt angle calculated by the forward tilt angle calculation unit.

3. the parameter value calculation unit further includes a forward lean duration calculation unit that calculates a forward lean duration for which the robot should maintain the forward lean angle based on the severity; The robot control system according to claim 2 , wherein the movement control unit controls the bowing movement in accordance with the forward leaning duration calculated by the forward leaning duration calculation unit.

4. the parameter value calculation unit further includes a forward lean time calculation unit that calculates a time until the robot leans forward based on the severity level, 3. The robot control system according to claim 1, wherein the movement control unit controls the bowing movement in accordance with the forward leaning time calculated by the forward leaning time calculation unit.

5. a determination unit that determines whether the apology accompanied by the bowing motion has been accepted; and a parameter value change unit that, when the determination unit determines that the apology has not been accepted, changes the parameter value to increase it for the same failure category; The robot control system according to claim 1 , wherein the movement control unit executes the bowing movement again in accordance with the parameter value changed by the parameter value change unit.

6. A method for controlling a robot that can perform a bowing motion by tilting at least one of its upper body and head forward, comprising: When the robot fails, determining a severity of the failure in a category; calculating a parameter value for controlling the bowing motion based on the determined severity; and A robot control method, further comprising: controlling the bowing motion in accordance with the calculated parameter value.

7. A robot control program for controlling a robot capable of bowing by tilting at least one of its upper body and head forward, comprising: The robot control program causes the computer of the robot to: a severity determination unit that, when the robot fails, determines the severity of the failure in a category; a parameter value calculation unit that calculates a parameter value for controlling the bowing action based on the severity determined by the severity determination unit; and a motion control unit that controls the bowing motion of the robot in accordance with the forward lean parameter value calculated by the parameter value calculation unit; A robot control program that functions as a

8. A robot capable of bowing by tilting at least one of its upper body and head forward, a severity determination unit that, when the robot fails, determines the severity of the failure in a category; a parameter value calculation unit that calculates a parameter value for controlling the bowing action based on the severity determined by the severity determination unit; and A robot comprising: a movement control unit that controls a bowing movement of the robot in accordance with the forward lean parameter value calculated by the parameter value calculation unit.

Images