Control device and request processing system provided with the same

Abstract

To provide a control device capable of generating a control signal for causing a processing unit to perform processing corresponding to a request while fundamental information and episode information effective for controlling the processing unit are unavailable, and a request processing system provided with the control device.SOLUTION: In a control device 20 that applies a control signal for processing to be carried out in response to a request to a processing unit 11 so as to exercise control, there are provided an episode storage part 21 for storing episode information formed by combining one or more out of visual information, auditory information, smell information, taste information, and tactile information detected by the processing unit 11 with position information and time information, and a control signal generation part 12 for generating a control signal. The control signal generation part 12 selects whether information used for generation of the control signal is set to one or both of the episode information stored by the episode storage part 21 and general information acquired from a large-scale language model 23.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a control device that generates a control signal for causing a processing unit to perform processing in response to a request, and to a request processing system equipped with the control device. [Background technology]

[0002] In a home service robot system in which a robot performs processing according to a request, information specific to the home being used, such as the layout of the site of use and the arrangement of objects, is preregistered in a control device that generates control signals for the robot. If the information registered in the control device is used as basic information and the control device uses only the basic information when generating control signals, a request for which usable basic information is not registered cannot be processed until the user updates the basic information.

[0003] In this regard, if the hippocampus model shown in Non-Patent Documents 1 and 2 is applied to a home service robot system, episode information resulting from the experience of using the robot at the site can also be used when generating a control signal. Here, episode information is composed of one or more of visual information, auditory information, olfactory information, taste information, and tactile information detected by the robot, combined with positional information and time information, such as information that a plastic bottle of mineral water was found in the refrigerator at 1:00 p.m. last Tuesday. [Prior art documents] [Non-patent literature]

[0004] [Non-Patent Document 1] Akinobu Mizutani, Yuichiro Tanaka, Hakaru Tamukoh, Katsumi Tateno, Osamu Nomura, and Takashi Morie, “Multiple memory accumulation and recall by a hippocampus model reflecting emotional values,” The 11th RIEC International Symposium on Brain Functions and Brain Computer (BFBC 2023), P-18, Sendai, February 17-18 (18), 2023. [Non-Patent Document 2] A. Mizutani, I. Kawashima, Y. Tanaka, K. Tamukai, K. Tatsuno, O. Nomura, T. Morie, “Behavior Generation of Home Service Robots Based on Experience Accumulation and Recollection Using a Hippocampal Model,” Neurocomputing Research Meeting (NC) of the Institute of Electronics, Information and Communication Engineers, NC2022-59, December 3, 2022, Osaka, Osaka Electro-Communication University. Summary of the Invention [Problem to be solved by the invention]

[0005] However, even in a home service robot system to which the hippocampus model is applied, if there is no information in the basic information and episode information that is effective for generating a control signal, it is not possible to make the robot perform processing according to a request. This is not limited to home service robot systems, but is also true for systems that have processing units other than a robot (for example, a monitor that displays content suited to a user in response to the user's request).

[0006] The present invention has been made in consideration of the above circumstances, and aims to provide a control device and a request processing system equipped with the same that are capable of generating control signals for causing a processing unit to perform processing in accordance with a request even without basic information and episode information that are effective for controlling the processing unit. [Means for solving the problem]

[0007] A control device according to a first invention that meets the above-mentioned objective is a control device that controls a processing unit by providing the processing unit with a control signal for causing the processing unit to perform processing according to a request, and is equipped with an episode memory unit that stores episode information formed by combining one or more of visual information, auditory information, olfactory information, gustatory information, and tactile information detected by the processing unit with positional information and temporal information, and a control signal generation unit that generates the control signal, and the control signal generation unit selects whether the information to be used in generating the control signal is the episode information stored in the episode memory unit or general information obtained from a large-scale language model, or both.

[0008] A second invention in line with the above-mentioned object provides a request processing system having a processing unit that performs processing in accordance with a request and a control signal generation unit that generates a control signal for causing the processing in accordance with the request to be performed and provides the control signal to the processing unit, and further comprising an episode memory unit that stores episode information formed by combining one or more of visual information, auditory information, olfactory information, gustatory information, and tactile information detected by the processing unit with positional information and temporal information, and the control signal generation unit selects whether the information to be used in generating the control signal is the episode information stored in the episode memory unit or general information obtained from a large-scale language model, or both. Effect of the Invention

[0009] The control device of the first invention comprises an episode memory unit that stores episode information formed by combining one or more of visual information, auditory information, olfactory information, gustatory information, and tactile information detected by a processing unit with positional information and temporal information, and a control signal generation unit that generates a control signal, and the control signal generation unit selects whether the information to be used in generating the control signal is the episode information stored in the episode memory unit or general information obtained from a large-scale language model, or both.

[0010] In addition, the request processing system of the second invention includes an episode memory unit that stores episode information formed by combining one or more of the visual information, auditory information, olfactory information, gustatory information, and tactile information detected by the processing unit with positional information and temporal information, and the control signal generation unit selects whether the information to be used in generating the control signal is the episode information stored in the episode memory unit or general information obtained from a large-scale language model, or both. Therefore, the control device according to the first aspect of the invention and the request processing system according to the second aspect of the invention are capable of generating a control signal even without basic information and episode information effective for controlling the processing unit. [Brief description of the drawings]

[0011] [Figure 1] FIG. 1 is an explanatory diagram of a request processing system according to a first embodiment of the present invention. [Diagram 2] FIG. 11 is an explanatory diagram of a request processing system according to a second embodiment of the present invention. [Diagram 3] FIG. 11 is an explanatory diagram of a request processing system according to a third embodiment of the present invention. [Figure 4] FIG. 13 is an explanatory diagram of a process flow of a request processing system according to a third embodiment of the present invention. [Diagram 5] FIG. 13 is an explanatory diagram of a process flow of a request processing system according to a third embodiment of the present invention. [Figure 6] FIG. 13 is an explanatory diagram of a process flow of a request processing system according to a third embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] Next, with reference to the attached drawings, an embodiment of the present invention will be described for understanding the present invention. 1, a request processing system 10 according to a first embodiment of the present invention is a system having a robot 11, which is an example of a processing unit that performs processing according to a request, and a control signal generating unit 12 that generates a control signal for causing the robot 11 to perform the processing according to the request. A detailed description will be given below.

[0013] 1, the robot 11 has an arm 13, a self-propelled mechanism 14, an imaging means 15, a plurality of sensors 16, a memory 17, a control unit 18, an input unit 19, etc. The arm 13 and the self-propelled mechanism 14 each have a plurality of drive units (motors, etc.), and the control unit 18 controls the arm 13 and the self-propelled mechanism 14 by sending command signals to operate these drive units. The arm 13 is designed to be able to grasp an object, and the self-propelled mechanism 14 is designed to allow the robot 11 to move.

[0014] The imaging means 15 has its imaging direction and the like adjusted so that it can capture images of the surroundings of the robot 11. The multiple sensors 16 are pressure sensors, distance sensors, and the like, and the control unit 18 can acquire the images captured by the imaging means 15 and the detection values ​​of each sensor 16 from the imaging means 15 and each sensor 16. The images captured by the imaging means 15 and the detection values ​​of each sensor 16 are stored in the memory 17 as appropriate.

[0015] The input unit 19 has one or more of a touch panel, a keyboard, and a microphone, and the user of the request processing system 10 can provide the request to the robot 11 (for example, a request to search the house for a TV remote control and bring it to the user) and information necessary for the robot 11 to perform processing (hereinafter also referred to as "basic information") to the request processing system 10 by operating or speaking to the input unit 19. The input unit 19 only needs to be capable of transmitting and receiving signals to and from the control unit 18, and may be, for example, a remote control with a wireless communication function.

[0016] The request processing system 10 includes a control device 20 that can be configured with a CPU, a memory, etc., and is connected by wire or wirelessly to the robot 11. In addition to a control signal generating unit 12, the control device 20 includes an episode storage unit 21 that stores episode information generated based on the behavior of the robot 11 (processing performed by the robot 11), etc., and a basic information storage unit 22 that stores basic information, etc.

[0017] Episode information is information that combines one or more of the visual information, auditory information, olfactory information, gustatory information, and tactile information (hereinafter collectively referred to as "perceptual information") detected by the sensors 16 of the robot 11, with information on the location where the perceptual information was detected (hereinafter also referred to as "location information") and information on the time (date and time) when the perceptual information was detected (hereinafter also referred to as "time information").

[0018] For example, in response to a request to search for a television remote control in the house and bring it to the user, if the robot 11 finds and acquires the television remote control on the dining table in the kitchen at 1:10 p.m. on November 11th, but finds it on the living table at 1:10 p.m. on the same day, the episode information that the television remote control was not on the low table in the living room at 1:05 p.m. on November 11th and that the television remote control was on the dining table in the kitchen at 1:10 p.m. on November 11th is generated.

[0019] The basic information is information about the home (site) where the request processing system 10 is used, such as the layout of the home (room layout, what is placed in which position in which room, etc.) The basic information is provided to the basic information storage unit 22 by the control device 20 acquiring it from outside via communication during the preparation stage before using the request processing system 10, and is stored in the basic information storage unit 22.

[0020] When the situation at home changes (for example, the layout is changed), new basic information is provided to the basic information storage unit 22, or the stored basic information is modified as appropriate. In this embodiment, the basic information storage unit 22 and the episode storage unit 21 are connected, and predetermined episode information stored in the episode storage unit 21 (for example, episode information in which the same event is detected a predetermined number of times within a predetermined period) is also stored in the basic information storage unit 22 as basic information.

[0021] In addition, the control device 20 is connected to a large-scale language model 23 via an external network. In this embodiment, the large-scale language model 23 is a model that collects information via an Internet line and returns information corresponding to a matter for which an answer is requested. The type of the large-scale language model 23 is not particularly limited. The control device 20 includes a data input / output unit 24 that transmits and receives information to and from the large-scale language model 23, and an analysis unit 25 that analyzes information provided from the large-scale language model 23 via the data input / output unit 24 (converts the information into information that can be used by the control signal generation unit 12) and provides the information to the control signal generation unit 12. In this embodiment, the control device 20 is configured as a chip, but is not limited to this.

[0022] The information given by the large-scale language model 23 obtained by the control signal generating unit 12 from the analyzing unit 25 is general information (common information) and not information specific to the home in which the request processing system 10 is used. For example, the information that the locations where the TV remote control is likely to be placed in an average home in Japan are on the TV stand, on the living room table, on the dining table, and on the sofa in the living room corresponds to general information.

[0023] When a new request is made to the robot 11 by the user, the control signal generating unit 12 obtains the necessary basic information from the basic information storing unit 22, and further detects the presence or absence of episode information available for generating a control signal corresponding to the new request in the episode storing unit 21. When the control signal generating unit 12 detects that episode information available for generating a control signal is available in the episode storing unit 21, it generates a control signal based on the basic information and the available information.

[0024] The term “episode information effective for generating a control signal” as used herein refers to one or more pieces of episode information stored in basic information storage unit 22, in the case where a control signal can be generated only from that one or more pieces of episode information.

[0025] On the other hand, when the control signal generation unit 12 detects that there is no episode information valid for generating a control signal in the episode storage unit 21, it requests general information related to the new request from the large-scale language model 23 via the data input / output unit 24. As a result, the control signal generation unit 12 obtains the general information from the large-scale language model 23 via the data input / output unit 24 and the analysis unit 25, and generates a control signal based on the basic information and the obtained general information. Here, there are two options for the control signal generator 12 to request general information related to a new request from the large-scale language model 23:

[0026] Option 1: Only new requests are queried into the large language model 23 . For example, if the new requirement is to find and bring back dry cell batteries, the inquiry for the new requirement will be where in an average household dry cell batteries might be placed.

[0027] Option 2: When a new request is queried from the large scale language model 23 , episode information stored in the episode store 21 is provided to the large scale language model 23 . For example, if a new request is to find and bring back batteries, the episode information provided to the large-scale language model 23 is information that the stock of light bulbs was on a shelf in the living room last month and information that the tools are stored in a closet in the bedroom. The episode information provided to the large-scale language model 23 may be one or more.

[0028] The information provided from the large-scale language model 23 differs depending on whether option 1 or 2 is selected, or, even if option 2 is selected, depending on the type of episode information provided to the large-scale language model 23. Whether option 1 or 2 is selected can be determined by settings.

[0029] Therefore, in this embodiment, if episode information corresponding to a request to the robot 11 is stored in the episode memory unit 21, a control signal is generated based on the episode information and basic information (using the relevant episode information), and if episode information corresponding to a request to the robot 11 is not stored in the episode memory unit 21, corresponding general information is obtained from the large-scale language model 23, and a control signal is generated based on the general information and basic information (using the relevant general information).

[0030] Furthermore, even if episode information available for generating a control signal exists in the episode storage unit 21, general information may be acquired from the large-scale language model 23, and a control signal may be generated based on the basic information, episode information, and general information. For example, if episode information available for generating a control signal exists in the episode storage unit 21, but the episode information in question is old (for example, episode information from six months ago), a control signal is generated based on the basic information, episode information, and general information. In this case, too, when the control signal generation unit 12 requests information from the large-scale language model 23, either of the above-mentioned options 1 and 2 can be adopted.

[0031] Therefore, the control signal generating unit 12 can select, according to a preset condition, whether the information to be used for generating a control signal corresponding to a new request is the episode information stored in the episode storage unit 21 or the general information acquired from the large-scale language model 23, or both, and generate a control signal using the selected information. Therefore, even if there is no episode information available for generating a control signal, the control signal generating unit 12 can generate a control signal using the general information acquired from the large-scale language model 23.

[0032] In the robot 11, the control unit 18 receives the control signal generated by the control signal generation unit 12, and operates the arm 13, the self-propelled mechanism 14, etc. to process the given request. If the request is fulfilled by the processing performed by the robot 11 (for example, the robot is able to search for the TV remote control and hand it over to the user), information generated during the process, such as the detection value of the sensor 16, is organized by the control unit 18 and sent to the episode storage unit 21, where it is added (stored) as new episode information. The robot 11 goes into a mode in which it waits for a new request.

[0033] On the other hand, if the request cannot be realized by the processing performed by the robot 11 (e.g., the TV remote control is searched for but not found), information generated during the process, such as the detection value of the sensor 16, is organized by the control unit 18 and sent to the episode storage unit 21, and new episode information (e.g., information that the TV remote control was not found on the dining table or living room table that were searched for) is added to the episode storage unit 21. Then, in order to generate a new control signal (e.g., a control signal for searching for a TV remote control at another location), new general information is obtained from the large-scale language model 23, or the robot 11 requests the user to input necessary information. After that, the robot 11 is given the new control signal and performs processing based on the control signal.

[0034] In addition, in the request processing system 10, the control device 20 is separated from the robot body and placed on the outside, which allows the robot body to be made lighter and reduces battery consumption, etc. Furthermore, depending on the design, it is possible to control multiple robots 11 simultaneously or individually with one control device 20. In this case, it is preferable to provide an episode storage unit 21 and a basic information storage unit 22 for each of the multiple robots 11.

[0035] Furthermore, when the large-scale language model 23 is acquired via the Internet or the like, the connectivity can be stabilized by providing the control device 20 outside the robot body. That is, when the control device 20 is provided inside the robot body and, for example, the connection to the Internet is made wirelessly, the transmission and reception state may become unstable depending on the location of the robot. Therefore, particularly when connecting to the Internet wirelessly, by fixing the control device 20 in a location separate from the robot body, the connection status to the Internet etc. can be stabilized and reliability can be improved.

[0036] The control device 20 can be configured as a personal computer such as a notebook computer or a desktop computer, or a mobile communication terminal such as a smartphone or a tablet terminal, which can implement the functions of the control device 20 by a program. Alternatively, the control device 20 may be configured as a one-chip IC on a module substrate, or may be configured by combining multiple semiconductor ICs. Note that, in this embodiment, the control device 20 is provided outside the robot body (not shown) that houses each part of the robot 11, but all or at least a part of the control device 20 may be provided inside the robot body.

[0037] By providing the entire control device 20 on the robot body, the data communication components between the robot 11 and the control device 20 can be simplified, and an autonomously operable robot 11 can be provided. In addition, depending on the specifications and usage of the robot, it is possible to store part of the control device 20 in the robot body. For example, it is possible to arrange the analysis unit 25 and the data input / output unit 24, which control access to the large-scale language model 23, outside the robot body and store the other parts in the robot body.

[0038] The request processing system 10 described so far is a system that does not have a large-scale language model 23, but as shown in Figure 2, a request processing system 30 according to a second embodiment of the present invention that has a large-scale language model 31 may also be adopted. The large-scale language model 31 can be configured by a predetermined software and a server, and is provided in the control device 32 (the control device 32 is equipped with the large-scale language model 31). The large-scale language model 31 collects information on the Internet via internal lines and Internet lines, and generates general information.

[0039] In the request processing system 30, a control signal generating unit 33 and an analyzing unit 34 that analyzes general information obtained from the large-scale language model 23 and provides the information to the control signal generating unit 33 are connected to the large-scale language model 31. Note that in the request processing system 30, components similar to those in the request processing system 10 are denoted by the same reference numerals and detailed explanations thereof will be omitted.

[0040] Also, the control signal generating unit may be divided into an evaluation integration means for making inquiries of information required for the episode storage unit and the large-scale language model, and a signal generating means for generating a control signal. With reference to Figs. 3 to 6, a control device 41 in which the control signal generating unit is divided into a signal generating means 42 and an evaluation integration means 43, and a request processing system 40 according to a third embodiment of the present invention including the control device 41 and the robot 11 will be described. Note that in the request processing system 40, components similar to those in the request processing system 10 are denoted by the same reference numerals and detailed description thereof will be omitted.

[0041] 3, the control device 41 includes an evaluation integration means 43 connected to a signal generation means 42, and the signal generation means 42 and the evaluation integration means 43 can exchange information (data). The evaluation integration means 43 can acquire information of requests from users to the request processing system 40 (hereinafter, also referred to as "request information") via the signal generation means 42. Furthermore, the evaluation integration means 43 can exchange information with the episode storage unit 44 and the large-scale language model 45, and can inquire of the episode storage unit 44 and the large-scale language model 45 (or either one of them) and acquire episode information and general information corresponding to the request information according to the acquired request information.

[0042] Then, the evaluation integration means 43, which has acquired the episode information and general information, provides information necessary for the signal generation means 42 to generate a control signal based on that information. Although not shown in the figure, the control device 41 is provided with a basic information storage unit, an analysis unit, etc. A specific example of the process performed by the request processing system 40 for a request will now be described.

[0043] <First example> When a user requests that a specific object be brought to the user, as shown in Figure 4, (1) the evaluation integration means 43 queries the episode memory unit 44 for episode information necessary for the robot 11 to bring the object to the user. When the corresponding episode information is found, (2) the evaluation integration means 43 acquires the corresponding episode information from the episode storage unit 44. (3) The signal generation means 42 acquires necessary information from the evaluation integration means 43 to generate a control signal, and (4) transmits the control signal to the robot 11. The robot 11 that has received the control signal generates episode information based on the detection values ​​of the sensors 16 while the robot 11 is in action, and (5) the episode information is stored in the episode storage unit 44 via the evaluation integration means 43.

[0044] On the other hand, (1) when the episode storage unit 44 does not contain episode information corresponding to the request in response to an inquiry from the evaluation integration means 43 to the episode storage unit 44, (2) the evaluation integration means 43 inquires of the large-scale language model 45 about general information required for the robot 11 to bring a specified object to the user, as shown in Fig. 5, and (3) acquires the corresponding general information from the large-scale language model 45. When the evaluation integration means 43 requests information from the large-scale language model 45, it can adopt either option 1 or 2 described above.

[0045] (4) The signal generating means 42 acquires general information from the evaluation integrating means 43 as information necessary for generating a control signal, generates a control signal, and (5) transmits the control signal to the robot 11. Episode information is generated based on the detection values ​​of the sensors 16 while the robot 11 is in action, and (6) the episode information is stored in the episode storage unit 44 via the evaluation integrating means 43.

[0046] In addition, if the control device 32 is unable to obtain information useful for generating a control signal from either the episode memory unit 44 or the large-scale language model 45 in response to a query from the evaluation integration means 43, the control device 32 receives information such as the position of a specified object from the user, creates a control signal, and transmits it to the robot 11.

[0047] <Second example> In response to a request from a user to bring a specific object to the user, the episode storage unit 44 may have episode information for that object, but the episode information may not be appropriate for use in creating a control signal. For example, this may be the case when the location of the object changes depending on the season, and the season (e.g., summer) of the episode information for the object stored in the episode storage unit 44 is different from the current season (e.g., winter) (at the time of the request).

[0048] In this regard, when episode information corresponding to a request to the robot 11 (episode information that in winter, the user brought the relevant object) is stored in the episode memory unit 44, the evaluation integration means 43 detects this and provides the episode information to the signal generation means 42, and the signal generation means 42 generates a control signal using the episode information.

[0049] On the other hand, if episode information corresponding to a request to the robot 11 is not stored in the episode storage unit 44, the evaluation integration means 43 (1) queries the large-scale language model 45 for general information related to the request, and (2) acquires the general information from the large-scale language model 45, as shown in FIG. 6. In this case, the evaluation integration means 43 can also adopt either option 1 or 2 described above when requesting information from the large-scale language model 45. For a request to bring winter cafe au lait to a user, general information related to the request is, for example, information about what item corresponds to the expression variation of cafe au lait (e.g., information that the item corresponding to the expression variation of cafe au lait is coffee milk) or information about where an item similar to cafe au lait (e.g., coffee) is kept in the home in winter.

[0050] Then, (3) when the evaluation integration means 43 determines that the general information related to the request obtained from the large-scale language model 45 is appropriate for use in generating a control signal, it obtains appropriate episode information from the episode storage unit 44 based on the general information and provides the episode information to the signal generation means 42, (4) the signal generation means 42 generates a control signal and transmits it to the robot 11. (5) The episode information generated by the subsequent actions of the robot 11 is stored in the episode storage unit 44.

[0051] For example, if the evaluation integration means 43 acquires information that the item that corresponds to the expression variation of cafe au lait is coffee milk, and episode information that coffee milk was found in the household in winter exists in the episode memory unit 44, the evaluation integration means 43 determines that the relevant general information acquired from the large-scale language model 45 is appropriate for use in generating a control signal, and provides the corresponding episode information to the signal generation means 42.

[0052] On the other hand, if the evaluation and integration means 43 determines that the general information related to the request obtained from the large-scale language model 45 is not appropriate for use in generating a control signal (for example, if the episode information that coffee milk was found in the household in winter is not stored in the episode memory unit 44), the control device 41 asks the user to input search destination candidates for the specified object.

[0053] <Third Example> If multiple different requests are made to the robot 11, some of the requests may be the same, but other parts may be different. For example, a request to fetch water for a family member in the morning and a request to fetch water for a visitor in the afternoon are the same in that they both involve fetching water, but differ in the time period for which they are requested and for whom. In response to this, the evaluation integration means 43 (i.e., the control signal generation unit) selects whether the episode storage unit 44 or the large-scale language model 45 is to be the destination to which to give priority for inquiries about information required for generating a control signal, depending on the content of the request, and makes an inquiry to the episode storage unit 44 or the large-scale language model 45 based on the selection.

[0054] Although the embodiment of the present invention has been described above, the present invention is not limited to the above-mentioned embodiment, and all changes in conditions that do not depart from the gist of the present invention are within the scope of application of the present invention. For example, a processing unit other than a robot (such as a unit that displays an appropriate response to an inquiry on a screen) may be used. [Explanation of symbols]

[0055] 10: Requirement processing system, 11: Robot, 12: Control signal generation unit, 13: Arm, 14: Self-propelled mechanism, 15: Imaging means, 16: Sensor, 17: Memory, 18: Control unit, 19: Input unit, 20: Control device, 21: Episode storage unit, 22: Basic information storage unit, 23: Large-scale language model, 24: Data input / output unit, 25: Analysis unit, 30: Requirement processing system, 31: Large-scale language model, 32: Control device, 33: Control signal generation unit, 34: Analysis unit, 40: Requirement processing system, 41: Control device, 42: Signal generation means, 43: Evaluation and integration means, 44: Episode storage unit, 45: Large-scale language model

Claims

1. A control device for controlling a processing unit by providing a control signal to the processing unit to cause the processing unit to perform processing in response to a request, an episode storage unit that stores episode information formed by combining one or more of visual information, auditory information, olfactory information, taste information, and tactile information detected by the processing unit with position information and time information; A control signal generating unit that generates the control signal, A control device characterized in that the control signal generation unit selects whether the information to be used in generating the control signal is the episode information stored in the episode memory unit, general information obtained from a large-scale language model, or both.

2. 2. The control device according to claim 1 , wherein, when the episode information corresponding to a request to the processing unit is stored in the episode storage unit, the control signal generation unit generates the control signal using the episode information, and when the episode information corresponding to the request to the processing unit is not stored in the episode storage unit, the control signal generation unit obtains the general information corresponding to the request to the processing unit from the large-scale language model and generates the control signal using the general information.

3. 2. The control device according to claim 1 , wherein, when the episode information corresponding to a request to the processing unit is stored in the episode storage unit, the control signal generation unit generates the control signal using the episode information, and when the episode information corresponding to the request to the processing unit is not stored in the episode storage unit, the control signal generation unit queries the large-scale language model to obtain general information related to the request to the processing unit, and determines the appropriateness of using the general information to generate the control signal.

4. 4. The control device according to claim 2 or 3, wherein when the control signal generation unit queries the large-scale language model to obtain the general information from the large-scale language model, the control signal generation unit provides the episode information stored in the episode storage unit to the large-scale language model.

5. 4. The control device according to claim 1, further comprising the large-scale language model.

6. A request processing system having a processing unit for performing processing in response to a request and a control signal generating unit for generating a control signal for causing the processing in response to the request to be performed and providing the control signal to the processing unit, A request processing system comprising an episode memory unit that stores episode information formed by combining one or more of visual information, auditory information, olfactory information, gustatory information, and tactile information detected by the processing unit with positional information and temporal information, and the control signal generation unit selects whether the information to be used in generating the control signal is the episode information stored in the episode memory unit or general information obtained from a large-scale language model, or both.

7. 7. The request processing system according to claim 6, wherein when the episode information corresponding to the request to the processing unit is stored in the episode storage unit, the control signal generation unit generates the control signal using the episode information, and when the episode information corresponding to the request to the processing unit is not stored in the episode storage unit, the control signal generation unit obtains the general information corresponding to the request to the processing unit from the large-scale language model and generates the control signal using the general information.

8. 7. The request processing system according to claim 6, wherein when the episode information corresponding to the request to the processing unit is stored in the episode storage unit, the control signal generation unit generates the control signal using the episode information, and when the episode information corresponding to the request to the processing unit is not stored in the episode storage unit, the control signal generation unit queries the large-scale language model to obtain general information related to the request to the processing unit, and determines the appropriateness of using the general information to generate the control signal.

9. 9. A request processing system according to claim 7 or 8, characterized in that when the control signal generation unit queries the large-scale language model to obtain the general information from the large-scale language model, the control signal generation unit provides the episode information stored in the episode storage unit to the large-scale language model.

10. 7. The request processing system according to claim 6, wherein the processing unit is a robot.

11. 9. A request processing system according to claim 6, further comprising said large-scale language model.

Images