Apparatus, method for determining action, and computer system

Abstract

To realize cooperative autonomous control by considering emotions of people or other devices.SOLUTION: A device holds an internal model that predicts a state of a future environment, acquires observation data indicating the state of the environment observed by the device, uses the observation data to generate first state information, uses the first state information to generate second state information of an intelligent agent existing in a periphery, determines a self-emotion of the device on the basis of the first state information, determines the self-emotion of the intelligent agent on the basis of the second state information, determines a social emotion on the basis of the self-emotions of the device and intelligent agent, generates future first state information and future second state information by using the internal model, and determines an action on the basis of the future first state information, future second state information, and social emotion.SELECTED DRAWING: Figure 3

Description

Technical Field

[0001] The present invention relates to an action control technology for an apparatus that performs autonomous control.

Background Art

[0002] Autonomous control technologies are being utilized in various industries such as the transportation industry, manufacturing industry, and nursing care industry. An apparatus equipped with an autonomous control technology acts to achieve a predetermined purpose according to the environment.

[0003] An apparatus that interacts with a person is required to perform control considering the mental state of the person existing in the environment.

[0004] For example, Patent Document 1 discloses a robot apparatus having a function of switching an action selection criterion considering its own state and an action selection criterion considering the state of others according to the situation.

Prior Art Documents

Patent Documents

[0005]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0006] In the technology described in Patent Document 1, each action selection criterion is independent. Therefore, cooperative autonomous control considering the emotions of a person or another apparatus cannot be realized. [[ID=D4]]

[0007] An object of the present invention is to realize cooperative autonomous control considering the emotions of a person or another apparatus.

Means for Solving the Problems

[0008] A representative example of the invention disclosed in this application is as follows: A device that exists in a space where an intelligent agent that acts based on a mental state exists and performs autonomous control, comprising a processor, a memory device connected to the processor, and an interface connected to the processor, which holds an internal model that predicts the future state of the environment based on the state of the environment, the social feelings that an individual has towards other individuals, and its behavior, and which performs the following processes via the interface: a process of acquiring observation data indicating the state of the environment observed by the device; a process of generating first state information using the observation data, which includes information about the intelligent agent present around the device as the state of the environment grasped by the device; and a process of generating second state information using the first state information, which includes information about the device and other intelligent agents present around the intelligent agent as the state of the environment grasped by the intelligent agent. The system then performs an action decision process which includes: a process for determining the self-emotions of the device based on the first state information; a process for determining the self-emotions of the intelligent agent based on the second state information; a process for determining the social emotions of the device towards the intelligent agent and the social emotions of the intelligent agent towards the device or other intelligent agents based on the self-emotions of the device and the self-emotions of the intelligent agent; a process for generating first predicted state information, which is a prediction of the future first state information, and second predicted state information, which is a prediction of the future second state information, using the internal model; and a process for determining the actions to be taken by the device based on the first predicted state information, the second predicted state information, the social emotions of the device towards the intelligent agent, and the social emotions of the intelligent agent towards the device or other intelligent agents. [Effects of the Invention]

[0009] According to the present invention, cooperative autonomous control that takes into account the emotions of a person or other device can be realized. Problems, configurations, and effects other than those described above will be clarified by the following description of embodiments. [Brief explanation of the drawing]

[0010] [Figure 1] It is a diagram showing the hardware configuration of the autonomous device of Example 1. [Figure 2] It is a diagram showing an example of the space where the autonomous device of Example 1 exists. [Figure 3] It is a diagram showing an example of the functional configuration of the autonomous device of Example 1. [Figure 4] It is a diagram showing an example of the internal model held by the autonomous device of Example 1. [Figure 5A] It is a diagram showing an example of the observation data database of Example 1. [Figure 5B] It is a diagram showing an example of the observation data database of Example 1. [Figure 6A] It is a diagram showing an example of the emotion data database of Example 1. [Figure 6B] It is a diagram showing an example of the emotion data database of Example 1. [Figure 7] It is a flowchart explaining an example of the process executed by the autonomous device of Example 1. [Figure 8A] It is a diagram showing an example of the emotion determination rule held by the autonomous device of Example 1. [Figure 8B] It is a diagram showing an example of the emotion determination rule held by the autonomous device of Example 1. [Figure 9] It is a diagram showing an example of the weight update rule held by the autonomous device of Example 1. [Figure 10A] It is a diagram showing an example of the display of the processing result output by the autonomous device of Example 1. [Figure 10B] It is a diagram showing an example of the display of the processing result output by the autonomous device of Example 1. [Figure 11] It is a diagram showing an example of the functional configuration of the autonomous device of Example 2.

Modes for Carrying Out the Invention

[0011] Hereinafter, embodiments of the present invention will be described with reference to the drawings. However, the present invention is not construed as being limited to the description of the embodiments shown below. Those skilled in the art can easily understand that the specific configuration can be changed without departing from the spirit or gist of the present invention.

[0012] In the configuration of the invention described below, the same or similar configurations or functions are denoted by the same reference numerals, and redundant descriptions are omitted.

[0013] The notations such as "first", "second", "third", etc. in this specification and the like are attached to identify the components, and do not necessarily limit the number or order.

[0014] The positions, sizes, shapes, and ranges of the respective configurations shown in the drawings and the like may not represent the actual positions, sizes, shapes, and ranges in order to facilitate the understanding of the invention. Therefore, in the present invention, it is not limited to the positions, sizes, shapes, and ranges disclosed in the drawings and the like.

Embodiment

[0015] FIG. 1 is a diagram showing the hardware configuration of the autonomous device of Embodiment 1. FIG. 2 is a diagram showing an example of the space in which the autonomous device of Embodiment 1 exists.

[0016] The autonomous device 100 is a robot, an automobile, etc., and acts based on inputs obtained from the environment and the emotions of people and other autonomous devices existing in the environment. The autonomous device 100 moves, for example, within the space 200 and performs operations such as transporting luggage. In the space 200, there are autonomous devices 201, 204 and people 202, 203. The autonomous device 100 acts considering the emotions of the autonomous devices 201, 204 and people 202, 203.

[0017] In this specification, an individual (person or autonomous device 100) that acts based on a mental state is described as an intelligent agent. The intelligent agents recognized by the autonomous device 100 are described as recognized intelligent agents.

[0018] The emotions of an intelligent agent include internal self-emotions and social emotions that an intelligent agent (an individual with emotions) feels towards other intelligent agents. Examples of self-emotions include fear and joy. Examples of social emotions include sympathy, trust, and resentment.

[0019] The autonomous device 100 includes a processor 101, memory 102, auxiliary storage device 103, network interface 104, observation device 105, control device group 106, input device 107, and output device 108. The hardware configuration of the autonomous device 100 is an example and is not limited thereto. For example, the autonomous device 100 may not have the input device 107 and the output device 108.

[0020] The processor 101 executes the program stored in memory 102. By executing processing according to the program, the processor 101 operates as a functional unit (module) that realizes a specific function. In the following description, when the processing is described with a functional unit as the subject, it indicates that the processor 101 is executing the program that realizes that functional unit.

[0021] Memory 102 stores the program executed by the processor 101 and the information that the program executes. Memory 102 is also used as a work area. The auxiliary storage device 103 is a large-capacity storage device that permanently stores data. The auxiliary storage device 103 is, for example, an HDD (Hard Disk Drive) and an SSD (Solid State Drive). The program and information stored in memory 102 may also be stored in the auxiliary storage device 103. In this case, the processor 101 reads the program and information from the auxiliary storage device 103 and loads it into memory 102.

[0022] The network interface 104 communicates with external devices via the network. The observation device 105 includes a camera, LiDAR, microphone, millimeter-wave radar, and acceleration sensor, and acquires observation data 350 (see Figure 3) containing information for understanding the state of the environment.

[0023] The control device group 106 includes a motor and a speech device, etc. The input device 107 includes a keyboard, mouse, and touch panel, etc., and accepts user input. The output device 108 includes a display and speaker, etc., and outputs information to the user.

[0024] Figure 3 shows an example of the functional configuration of the autonomous device 100 in Example 1. Figure 4 shows an example of the internal model held by the autonomous device 100 in Example 1.

[0025] The autonomous device 100 includes an intelligent agent detection / tracking unit 301, a state information generation unit 302, a free energy calculation unit 303, an emotion estimation unit 304, an action decision unit 305, and a learning unit 306. The autonomous device 100 also maintains an observation data database 311, an internal model database 312, and an emotion data database 313.

[0026] The autonomous device 100 acquires observation data 350 via the observation device 105. The observation data 350 is input to the intelligent agent detection / tracking unit 301 and the free energy calculation unit 303. The autonomous device 100 also receives action target information 351 via the network interface 104 or the input device 107.

[0027] Observational data 350 includes, for example, image and point cloud data. Behavioral target information 351 is information for controlling the selection of behavior.

[0028] The observation data database 311 is a database that manages observation data 350 and state information representing the state of the environment. The internal model database 312 is a database for managing the internal model in the free energy principle. The emotion data database 313 is a database for managing the self-emotions and social emotions of intelligent agents.

[0029] The intelligent agent detection / tracking unit 301 uses the observation data 350 to detect intelligent agents present around the autonomous device 100 and to track them. The intelligent agent detection / tracking unit 301 generates the coordinates and velocity of the detected intelligent agents as self-state information representing the state of the environment as perceived by the autonomous device 100. The intelligent agent detection / tracking unit 301 stores the observation data 350 and state information in the observation data database 311 and outputs the observation data 350 and processing results to the state information generation unit 302.

[0030] The state information generation unit 302 uses the observed data 350 and state information to generate state information (other state information) that represents the state of the environment as perceived by the intelligent recognition agent. For example, if coordinates and velocity are obtained as state information, other state information can be generated by transforming the coordinates of the state information. If the observed data 350 is an image, existing technologies such as Novel View Synthesis can be used to generate an image (other observed data) acquired by the intelligent agent, and other state information can be generated from this other observed data. The present invention is not limited to the algorithm for generating other state information.

[0031] In the following, when there is no distinction between self-state information and other-state information, it will be referred to simply as "state information."

[0032] The free energy calculation unit 303 calculates the free energy necessary to determine the self-emotions and social emotions of the intelligent agents (autonomous device 100 and cognitive intelligent agents). The free energy calculation unit 303 also calculates the expected free energy necessary to determine the actions of the intelligent agents (autonomous device 100 and cognitive intelligent agents).

[0033] The free energy is determined by an internal model Q, which is based on the latent state s of the environment, the generative model P(s,o) of the observed data o, and the parameter θ that approximates the generative model of the latent variable s of the environment. θ It is defined by equation (1) using (s) and the observed data o. KL[ ] is the Kullback-Leibler distance, and E[ ] is the expected value.

[0034]

number

[0035] Here, the internal model of Example 1 will be explained using Figure 4. Figure 4 is an example of an internal model in which a neural network is used to construct a latent state encoder and decoder. The internal model takes self-state, intelligent agent state, social emotion, and behavior as input and outputs a predicted self-state and a predicted intelligent agent state. Social emotion is input as binary data, a one-hot code, or a combination of a value obtained from the free energy of the self-intelligent agent and a value obtained from the free energy of the other intelligent agent. The self-state is the state of the self-intelligent agent included in the state information, and the intelligent agent state is the state of the other intelligent agent included in the state information. The encoder has a reparameterization trick, which is used in variational autoencoders (VAEs), and probabilistically samples the latent state. In other words, the internal model is a model that probabilistically outputs future state information from current state information. The internal model may also have parameters to control learning, such as in β-VAE.

[0036] The expected free energy is defined by equation (2). π represents the sequence of actions, and tilde s and tilde o are the sequence of latent state transitions and the sequence of observed data when the sequence of actions π is performed. C is a probability distribution representing the agent's preference for the observed data.

[0037]

number

[0038] In active reasoning, the expected free energy for each possible behavioral hypothesis is determined, and the probability distribution of the behavior is defined, for example, by the softmax function in equation (3), such that the smaller the expected free energy, the higher the probability of selecting that behavioral hypothesis.

[0039]

number

[0040] The free energy calculation unit 303 calculates the free energy of the autonomous device 100 using its own state information and internal model. The free energy calculation unit 303 also calculates the free energy of the intelligent agent using other state information and internal model. The calculation of the intelligent agent's free energy requires the intelligent agent's internal model. In Example 1, based on the simulation theory, which states that people empathize with others by simulating their sensations using their own bodies, the internal model of the intelligent agent is substituted with the internal model of the autonomous device 100.

[0041] The free energy calculation unit 303 generates future self-state information by inputting its own state information into an internal model, and calculates the expected free energy of the autonomous device 100 using the future self-state information. The free energy calculation unit 303 generates future other-state information by inputting other-state information into an internal model, and calculates the expected free energy of the recognizing intelligent agent using the future other-state information.

[0042] The emotion estimation unit 304 estimates the self-emotions of the autonomous device 100, the social emotions of the autonomous device 100 towards the intelligent agent, the self-emotions of the intelligent agent, and the social emotions of the intelligent agent towards the autonomous device 100 or other intelligent agents, based on the free energy of the autonomous device 100 and the recognizing intelligent agent. The estimation method will be described later.

[0043] The action decision unit 305 determines the action to be taken by the autonomous device 100 based on the expected free energy of the autonomous device 100 and the recognition intelligent agent, equation (3), and the action target information 351. The learning unit 306 learns an internal model.

[0044] Furthermore, regarding the functional units of the autonomous device 100, multiple functional units may be combined into one functional unit, or one functional unit may be divided into multiple functional units.

[0045] Figures 5A and 5B show an example of the observation data database 311 of Example 1. The observation data database 311 stores tables 500 and 510.

[0046] Table 500 is a table that stores self-state information generated by the intelligent agent detection / tracking unit 301. Table 500 includes entries consisting of intelligent agent ID 501, timestamp 502, position 503, and speed 504. There is one entry for each pair of intelligent agent and timestamp. One entry represents the state of the intelligent agent recognized by the autonomous device 100 at a certain time.

[0047] Intelligent Agent ID 501 is a field that stores the identification information of the detected intelligent agent. The state of the autonomous device 100 itself is also managed in table 500. Timestamp 502 is a field that stores the timestamp included in the observation data 350. Position 503 and velocity 504 are a group of fields that store values ​​representing the state understood by the autonomous device 100, calculated from the observation data 350.

[0048] Table 510 is a table that stores other state information generated by the state information generation unit 302. Table 510 includes entries consisting of an intelligent agent ID 511, a timestamp 512, a position 513, and a speed 514. There is one entry for each pair of intelligent agent and timestamp. One entry represents other state information for one recognized intelligent agent.

[0049] Intelligent Agent ID 511 is a field that stores the identification information of the detected intelligent agent. Note that there is no entry for the autonomous device 100 in table 510. Timestamp 512 is a field that stores the timestamp. Timestamp 512 stores the timestamp of the state information observation data 350 used to generate other state information. Position 513 and velocity 514 are a group of fields that store values ​​representing the state grasped by the recognizing intelligent agent.

[0050] Figures 6A and 6B show an example of the emotion data database 313 of Embodiment 1. The emotion data database 313 stores tables 600 and 610.

[0051] Table 600 is a table for managing self-emotions. Table 600 contains entries consisting of an intelligent agent ID 601, a timestamp 602, and a self-emotion 603. There is one entry for each combination of intelligent agent and timestamp.

[0052] Intelligent Agent ID 601 is a field that stores the identification information of the intelligent agent. Timestamp 602 is a field that stores the timestamp of the state information used to calculate the self-emotion. Self-emotion 603 is a field that stores the self-emotion of the intelligent agent.

[0053] Table 610 is a table for managing social emotions. There is a separate Table 610 for each intelligent agent. The Table 610 shown in Figure 6 is a table for managing social emotions towards the recognizing intelligent agent of the autonomous device 100.

[0054] Table 610 contains entries consisting of a timestamp 611, intelligent agent ID 612, location 613, speed 614, and social emotion 615. There is one entry for each combination of intelligent agent and timestamp.

[0055] The timestamp 611 is a field that stores the timestamp of the state information used to calculate the expected free energy. When multiple observation data 350 are input as time series data, the timestamp of the first or last observation data 350 of the time series data is stored in the timestamp 611. The intelligent agent ID 612 is a field that stores the identification information of the intelligent agent. The position 613 and velocity 614 are a group of fields that store predicted state information calculated using the internal model. The social sentiment 615 is a field that stores the social sentiment of the intelligent agent towards other intelligent agents.

[0056] Figure 7 is a flowchart illustrating an example of the processing performed by the autonomous device 100 in Example 1. Figures 8A and 8B show an example of the emotion determination rule held by the autonomous device 100 in Example 1. Figure 9 shows an example of the weight update rule held by the autonomous device 100 in Example 1.

[0057] The autonomous device 100 uses the received observation data 350 to perform the processing described below. The timing of the execution can be set arbitrarily. For example, it may be executed when the observation data 350 is received, or it may be executed when a predetermined number of observation data 350 has been accumulated.

[0058] The intelligent agent detection / tracking unit 301 uses the observed data 350 to track the intelligent agent detectionand perform tracking (step S101).

[0059] The state information generation unit 302 generates state information of the recognition intelligent agent using the observed data 350 (step S102).

[0060] The free energy calculation unit 303 starts loop processing of the intelligent agent (step S103). Specifically, the free energy calculation unit 303 selects one intelligent agent from among the autonomous device 100 and the recognition intelligent agent.

[0061] The free energy calculation unit 303 calculates the free energy using the state information of the intelligent agent (step S104).

[0062] The emotion estimation unit 304 determines the self-emotion based on free energy (step S105).

[0063] Here, we will explain the emotion determination rules managed by the autonomous device 100. The autonomous device 100 maintains tables 800 and 810 as emotion determination rules.

[0064] Table 800 is a table for managing the emotion determination rules for determining one's own emotions. Table 800 stores entries consisting of F'801 and self-emotion 802.

[0065] F'801 is free energy first derivative This field stores conditions related to the value of . Self-emotion 802 is a field that stores a value indicating self-emotion. In this embodiment, if the value of the first derivative of free energy is positive, self-emotion is determined to be negative, and if the value of the first derivative of free energy is negative, self-emotion is positive It is determined to be so.

[0066] The judgment rules shown in Table 800 are examples only and are not limited thereto. For example, self-emotion may be defined for combinations of conditions relating to the first derivative of free energy and conditions relating to the second derivative. This allows for the setting of a variety of self-emotions.

[0067] Table 810 is a table for managing the emotion determination rules for determining social emotions. Table 810 stores entries consisting of self-emotion (self) 811, self-emotion (other) 812, and social emotion 813.

[0068] Self-emotion (Self) 811 is a field that stores the self-emotion of an intelligent agent. Self-emotion (Other) 812 is a field that stores the self-emotion of an intelligent agent of another. Social emotion 813 is a field that stores a value indicating social emotion.

[0069] The emotion estimation unit 304 determines the intelligent agent's self-emotion using the intelligent agent's free energy and the table 800.

[0070] The free energy calculation unit 303 determines whether processing has been completed for all intelligent agents (step S106).

[0071] If processing is not complete for all intelligent agents, the free energy calculation unit 303 returns to step S103 and selects a new intelligent agent.

[0072] Once processing is complete for all intelligent agents, the emotion estimation unit 304 starts loop processing for the intelligent agents (step S107). Specifically, the emotion estimation unit 304 selects one intelligent agent from among the autonomous device 100 and the recognized intelligent agents. The selected intelligent agent is referred to as the target intelligent agent.

[0073] The emotion estimation unit 304 determines the social emotions of the target intelligent agent towards other intelligent agents (step S108). Specifically, the following processes are performed.

[0074] (S108-1) The emotion estimation unit 304 selects one intelligent agent from among the intelligent agents excluding the target intelligent agent. The selected intelligent agent is referred to as the sub-target intelligent agent.

[0075] (S108-2) The emotion estimation unit 304 determines the social emotion of the target intelligent agent towards the sub-target intelligent agent based on the target intelligent agent's own emotions, the sub-target intelligent agent's own emotions, and the table 810.

[0076] (S108-3) The emotion estimation unit 304 determines whether processing has been completed for all intelligent agents except the target intelligent agent. If processing has not been completed for all intelligent agents except the target intelligent agent, the emotion estimation unit 304 returns to S108-1 and selects a new sub-target intelligent agent. If processing has been completed for all intelligent agents except the target intelligent agent, the emotion estimation unit 304 terminates the processing in step S108.

[0077] The free energy calculation unit 303 generates future state information for the intelligent agent (step S109).

[0078] Specifically, the free energy calculation unit 303 retrieves the latest social emotion from the intelligent agent table 610 stored in the emotion data database 313. The free energy calculation unit 303 retrieves the intelligent agent's behavior from the hypothetical behavior decision unit 305. The free energy calculation unit 303 calculates future state information by inputting the intelligent agent's state information, social emotion, and behavior into an internal model.

[0079] The free energy calculation unit 303 uses future state information to calculate the expected free energy of the intelligent agent for a hypothetical action (step S110).

[0080] The emotion estimation unit 304 determines whether processing has been completed for all intelligent agents (step S111).

[0081] If processing is not complete for all intelligent agents, the emotion estimation unit 304 returns to step S107 and selects a new intelligent agent.

[0082] Once processing is complete for all intelligent agents, the action decision unit 305 determines the action of the autonomous device 100 based on the expected free energy and social emotions (step S112).

[0083] The action decision unit 305 of Example 1 determines action a based on the integrated expected free energy shown in equation (4) and the action target information 351. Here, index i represents the recognizing intelligent agent. my G is the expected free energy of the autonomous device 100, oth i W is the expected free energy of the cognitive intelligent agent. my and W oth i This represents weight.

[0084]

number

[0085] The action decision unit 305 adjusts the weights according to social emotions. For example, the weights are adjusted based on a table 900 as shown in Figure 9. Table 900 includes entries consisting of social emotions 901 and weight adjustment rules 902. Social emotions 901 is a field that stores social emotions. Weight adjustment rules 902 is a field that stores weight adjustment rules. The amount of change in weight is assumed to be predetermined.

[0086] The autonomous device 100 can output the results of the processing described in Figure 7 to an external source. Figures 10A and 10B show examples of the display of the processing results output by the autonomous device 100 of Embodiment 1.

[0087] The screen 1000 shown in Figure 10A includes icon description fields 1001 and 1002, and display fields 1003, 1004, and 1005.

[0088] Icon description fields 1001 and 1002 are fields for displaying explanations of icons representing personal and social emotions.

[0089] Display field 1003 is a field that displays the results of various processes based on the autonomous device 100. Display field 1003 displays the autonomous device 100's own emotions, the autonomous device 100's social emotions towards the cognitive intelligent agent, and the predicted results of the actions of the autonomous device 100 and the cognitive intelligent agent. Display field 1004 is a field that displays the results of various processes based on the cognitive intelligent agent specified by the user. Display field 1005 is a field that displays text representing the processing results.

[0090] Furthermore, free energy, predicted observational data, and weights may be displayed in display fields 1003 and 1004.

[0091] Furthermore, as shown in Figure 10B, screen 1000 may display information regarding the free energy used to determine emotions. For example, display area 1003 may display a graph showing the first and second derivative values ​​of the free energy of each intelligent agent. Display area 1004 may show the first derivative value of the free energy of a specified intelligent agent on the horizontal axis and the first derivative value of the free energy of other intelligent agents on the horizontal axis. Vertical axis A graph showing this will be displayed.

[0092] The learning unit 306 performs the learning process of the internal model using state information and social emotions. If the internal model is composed of an encoder and decoder using a neural network, social emotions are input as conditions to the encoder and decoder, and the learning process is executed. Furthermore, if the internal model has parameters that control learning, such as β-VAE, the parameters that control learning are set based on social emotions.

[0093] According to Example 1, the autonomous device 100 can control its actions based on its own emotions and social emotions towards a recognizable intelligent agent. In other words, it can achieve cooperative autonomous control that takes into account the emotions of a person or other device.

[0094] (Modified Version) A computer system that centrally manages the autonomous device 100 may maintain the functional configuration shown in Figure 3, acquire observation data 350 from the autonomous device 100, and perform calculation of free energy, estimation of emotions, and determination of actions. [Examples]

[0095] In Example 2, the autonomous device 100 differs in that it transmits and receives information with other autonomous devices via communication. Below, Example 2 will be described focusing on the differences from Example 1.

[0096] The hardware configuration of the autonomous device 100 in Example 2 is the same as that of Example 1. The functional configuration of the autonomous device 100 in Example 2 differs in some respects from that of Example 1. Figure 11 shows an example of the functional configuration of the autonomous device 100 in Example 2.

[0097] The autonomous device 100 of Embodiment 2 newly includes a communication unit 307. The other functional configurations are the same as those of Embodiment 1. The communication unit 307 transmits and receives information between autonomous devices 100. The communication unit 307 is connected to the intelligent agent detection / tracking unit 301, the state information generation unit 302, the free energy calculation unit 303, and the emotion estimation unit 304, and outputs information according to the type of information received from other autonomous devices 100. The communication unit 307 also acquires information from any of the intelligent agent detection / tracking unit 301, the state information generation unit 302, the free energy calculation unit 303, and the emotion estimation unit 304 and transmits it to other autonomous devices 100.

[0098] The data structure of each database held by the autonomous device 100 in Example 2 is the same as that in Example 1.

[0099] The processing performed by the autonomous device 100 in Example 2 is the same as in Example 1. However, in each processing step, if there is information received from another autonomous device 100, that information is used.

[0100] By sending and receiving information between autonomous devices 100, it becomes possible to identify intelligent agents that were not detected by one autonomous device 100 due to shielding. Furthermore, a reduction in the computational load on the autonomous devices 100 is expected. Additionally, more coordinated control between the autonomous devices 100 can be achieved.

[0101] It should be noted that the present invention is not limited to the embodiments described above, and various modifications are included. Furthermore, for example, the embodiments described above are detailed explanations of the configuration in order to clearly illustrate the present invention, and are not necessarily limited to those having all the configurations described. In addition, some of the configurations in each embodiment can be added to, deleted from, or replaced with other configurations.

[0102] Furthermore, each of the above-mentioned configurations, functions, processing units, processing means, etc., may be implemented in hardware, in whole or in part, for example, by designing them as integrated circuits. The present invention can also be implemented by software program code that realizes the functions of the embodiment. In this case, a storage medium on which the program code is recorded is provided to a computer, and the processor of that computer reads the program code stored in the storage medium. In this case, the program code read from the storage medium itself realizes the functions of the embodiment described above, and the program code itself and the storage medium on which it is stored constitute the present invention. Examples of storage media used to supply such program code include flexible disks, CD-ROMs, DVD-ROMs, hard disks, SSDs (Solid State Drives), optical disks, magneto-optical disks, CD-Rs, magnetic tapes, non-volatile memory cards, ROMs, and the like.

[0103] Furthermore, the program code that implements the functions described in this embodiment can be implemented in a wide range of programming or scripting languages, such as assembler, C / C++, Perl, Shell, PHP, Python, and Java (registered trademark).

[0104] Furthermore, the program code for the software that implements the functions of the embodiment may be distributed via a network and stored in a storage means such as a computer's hard disk or memory, or in a storage medium such as a CD-RW or CD-R, and the computer's processor may read and execute the program code stored in the storage means or storage medium.

[0105] In the above-described embodiment, the control lines and information lines shown are those deemed necessary for explanation and do not necessarily represent all control lines and information lines in the actual product. All components may be interconnected. [Explanation of symbols]

[0106] 100 Autonomous Devices 101 Processors 102 memory 103 Auxiliary storage device 104 Network Interfaces 105 Observation equipment 106 Control device group 107 Input device 108 Output device 200 space 201 Autonomous device 202 people 301 Intelligent Agent Detection / Tracking Unit 302 State Information Generation Unit 303 Free Energy Calculation Unit 304 Emotion estimation section 305 Decision-Making Department 306 Learning Department 307 Liaison Department 311 Observational Data Database 312 Internal Model Database 313 Emotional Data Database 350 observation data 351 Action Goal Information 1000 screens

Claims

1. A device that exists in a space where intelligent agents that act based on mental states exist, and which performs autonomous control, The system comprises a processor, a storage device connected to the processor, and an interface connected to the processor. It maintains an internal model that predicts future environmental conditions based on the state of the environment, the social feelings individuals have towards other individuals, and their behaviors. The process of acquiring observation data indicating the state of the environment observed by the device via the interface, Using the aforementioned observation data, a process is performed to generate first state information, which includes information about the intelligent agents present around the device, as the state of the environment as perceived by the device. Using the first state information, a process is performed to generate second state information, which includes information about the devices and other intelligent agents present around the intelligent agent, as the state of the environment as perceived by the intelligent agent. A process to determine the self-emotion of the device based on the first state information, Based on the second state information, a process is performed to determine the self-emotions of the intelligent agent, A process for determining the social feelings of the device toward the intelligent agent and the social feelings of the agent toward the device or other intelligent agents, based on the self-feelings of the device and the self-feelings of the intelligent agent. A process that uses the aforementioned internal model to generate first predicted state information, which is a prediction of the future first state information, and second predicted state information, which is a prediction of the future second state information. An apparatus characterized by performing an action decision process that includes a process for determining an action to be taken by the apparatus based on the first predicted state information, the second predicted state information, the apparatus's social feelings toward the intelligent agent, and the intelligent agent's social feelings toward the apparatus or other intelligent agents.

2. The apparatus according to claim 1, Using the first state information, the free energy of the device according to the free energy principle is calculated, and based on the free energy of the device, the self-emotion of the device is determined. Using the second state information, the free energy of the intelligent agent is calculated, and based on the free energy of the intelligent agent, the self-emotion of the intelligent agent is determined. A device characterized by determining the device's social feelings toward the intelligent agent and the intelligent agent's social feelings toward the device or other intelligent agents, based on a combination of the device's own feelings and the intelligent agent's own feelings.

3. The apparatus according to claim 2, Using the first predicted state information, the expected free energy of the device under the free energy principle is calculated. Using the second predicted state information, the expected free energy of the intelligent agent is calculated. Based on the weighted sum of the expected free energy of the device and the expected free energy of the intelligent agent, the behavior of the device is determined. The apparatus is characterized in that the weights multiplied by the expected free energy of the apparatus and the expected free energy of the intelligent agent are determined based on the apparatus's social feelings toward the intelligent agent and the intelligent agent's social feelings toward the apparatus or other intelligent agents.

4. The apparatus according to claim 1, The intelligent agent includes other devices that perform the action decision processing. The aforementioned device acquires the calculation results calculated in the action decision process from the other device, An apparatus characterized by performing the action decision process using the first state information and the acquired calculation result.

5. The apparatus according to claim 1, The learning process to train the aforementioned internal model is executed, The apparatus is characterized in that the learning process uses the social feelings of the apparatus toward the intelligent agent and the social feelings of the intelligent agent toward the apparatus or other intelligent agents.

6. A method for determining an action that is performed by a device that performs autonomous control in a space where intelligent agents that act based on mental states exist, The aforementioned device is The system comprises a processor, a storage device connected to the processor, and an interface connected to the processor. It maintains an internal model that predicts future environmental conditions based on the state of the environment, the social feelings individuals have towards other individuals, and their behaviors. The aforementioned method for determining action is: The first step is for the processor to acquire observation data indicating the state of the environment observed by the device via the interface, The processor, using the observed data, generates first state information which includes information about the intelligent agents present around the device as the state of the environment as perceived by the device; A third step in which the processor uses the first state information to generate second state information, which includes information about the devices and other intelligent agents present around the intelligent agent, as the state of the environment as perceived by the intelligent agent. The processor performs a fourth step of determining the self-emotion of the device based on the first state information, A fifth step in which the processor determines the self-emotion of the intelligent agent based on the second state information, A sixth step in which the processor determines, based on the self-emotions of the device and the self-emotions of the intelligent agent, the social emotions of the device toward the intelligent agent and the social emotions of the intelligent agent toward the device or other intelligent agents. A seventh step in which the processor uses the internal model to generate first predicted state information, which is a prediction of the future first state information, and second predicted state information, which is a prediction of the future second state information. An action determination method characterized in that the processor determines an action to be taken by the device based on the first predicted state information, the second predicted state information, the device's social feelings toward the intelligent agent, and the intelligent agent's social feelings toward the device or other intelligent agents.

7. A method for determining an action according to claim 6, The fourth step described above is: The processor performs the step of calculating the free energy of the device in the free energy principle using the first state information, The processor includes the step of determining the self-feel of the device based on the free energy of the device, The fifth step described above is: The processor performs the step of calculating the free energy of the intelligent agent using the second state information, The processor includes the step of determining the self-emotion of the intelligent agent based on the free energy of the intelligent agent, The sixth step of the behavior determination method is characterized in that the processor determines the social feelings of the device toward the intelligent agent and the social feelings of the intelligent agent toward the device or other intelligent agents, based on a combination of the device's own feelings and the intelligent agent's own feelings.

8. A method for determining an action according to claim 7, The eighth step described above is: The processor performs the step of calculating the expected free energy of the device in the free energy principle using the first predicted state information, The processor performs the step of calculating the expected free energy of the intelligent agent using the second predicted state information, The processor includes the step of determining the behavior of the device based on a weighted sum of the expected free energy of the device and the expected free energy of the intelligent agent, A method for determining behavior, characterized in that the weights multiplied by the expected free energy of the device and the expected free energy of the intelligent agent are determined based on the social feelings of the device toward the intelligent agent and the social feelings of the intelligent agent toward the device or other intelligent agents.

9. A computer system that exists in a space where intelligent agents that act based on mental states exist, and is connected to a device that performs autonomous control, It maintains an internal model that predicts future environmental conditions based on the state of the environment, the social feelings individuals have towards other individuals, and their behaviors. The aforementioned device acquires observational data indicating the state of the environment observed, Using the aforementioned observation data, the device generates first state information, which includes information about the intelligent agents present around the device, as the state of the environment as perceived by the device. Using the first state information, a second state information is generated, which includes information about the devices and other intelligent agents present around the intelligent agent, as the state of the environment as perceived by the intelligent agent. Based on the first state information, the self-emotion of the device is determined, Based on the second state information, the self-emotions of the intelligent agent are determined. Based on the self-emotions of the device and the self-emotions of the intelligent agent, the social emotions of the device towards the intelligent agent and the social emotions of the intelligent agent towards the device or other intelligent agents are determined. Using the aforementioned internal model, a first predicted state information, which is a prediction of the future first state information, and a second predicted state information, which is a prediction of the future second state information, are generated. A computer system characterized in that it determines the actions to be taken by the device based on the first predicted state information, the second predicted state information, the device's social feelings toward the intelligent agent, and the intelligent agent's social feelings toward the device or other intelligent agents.

Images