JP2009151634A - Inconsistent process detecting apparatus, inconsistent process detecting program, and inconsistent process detecting method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an inconsistent process detecting apparatus, an inconsistent process detecting program, and an inconsistent process detecting method for quickly detecting inconsistent processes due to an agent. <P>SOLUTION: A monitoring unit 3 monitors a memory unit 7, which stores simulation results by an agent controller 11 or a process history by an agent simulator 100; and when an agent is detected which matches with inconsistent process detecting rule set 4, the monitoring unit 3 stores the agent states detected and states of agent simulator 100 at the detected time into an instance memory unit 5. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an inconsistency processing detection apparatus, an inconsistency processing detection program, and an inconsistency processing detection method for detecting inconsistent processing of agents in a rule-based simulator.

  There are simulators for simulating system processing and supporting system development and design. These simulators include agent simulators that simulate social systems.

An agent is software that realizes an action subject with autonomy. A system that operates in cooperation with multiple agents is called a multi-agent simulator.

  In an agent simulator, a rule base is often used to model system operation. Each rule (control rule) is considered to be independent, and a description of a format representing “then” action “else” action is often used from the “if” condition. Each rule is written declaratively and evaluated and processed independently. Each agent performs actions based on these control rules, and they interact to enable simulation of complex systems.

  In the case of a simulator using a rule base, it is necessary to create an appropriate rule set to simulate the system, but it is very difficult to predict in advance how those rules will fire.

In rule-based system development, the task of building rules by the developer is the most important and requires work time. Therefore, there are debugging technologies that support developers. A developer sets breakpoints based on selection conditions that occur during the execution of a program that employs a rule-based system, and when the program reaches a breakpoint, the developer checks the state of the system and changes the rules. This is a debugging method (for example, Patent Document 1).

However, with these methods, the program must be run repeatedly, and every time an inconsistent operation occurs and a breakpoint is reached, the developer must debug and evaluate the correction results, which is very There is a problem that it takes time. In particular, when a rule-based simulator system having a large number of agents is created, much more debugging time is required.
Japanese Patent Laid-Open No. 5-151025

  An object of the present invention is to provide an inconsistent operation detecting apparatus, method, and program capable of easily detecting inconsistent processing occurring in an execution result of an agent simulator.

  In order to solve the above-mentioned problem, the invention of claim 1 is characterized in that an initial state generation unit that gives an initial state in a virtual environment, a first storage unit that stores a control rule for controlling processing of an agent operation, An internal state generation unit that generates an internal state of the agent generated by giving a state in a virtual environment to the control rule, and a first storage that stores the internal state of the agent generated by the internal state generation unit A mismatch processing detection device for detecting inconsistent processing as processing for operating an agent that is a virtual action subject in a virtual environment constructed by a simulator device comprising: Is a second storage unit that stores a detection rule for detecting control processing of the agent that is regarded as inconsistent, and the agent of the simulator device A monitoring unit that monitors the control processing of the inconsistent agent corresponding to the detection rule, and when the monitoring unit detects the inconsistent processing, the internal state of the agent is controlled to be inconsistent. There is provided an inconsistency processing detection device characterized in that a second storage unit is provided for storing as a processed case.

  The invention according to claim 7 provides a function for generating an initial state in a virtual environment, a function for storing a control rule for controlling processing of an agent operation, and a state in the virtual environment for the control rule. And a function for generating the internal state of the agent and a function for storing the generated internal state of the agent. A program for detecting a consistent process, a function for storing a detection rule for detecting the control process of the agent, which is regarded as inconsistent as the process of the simulator, and the detection of the agent process of the simulator device A function to monitor the control processing of the inconsistent agent corresponding to the rule, and corresponding to the detection rule When a matching control process of the agent is detected, the computer realizes a function of storing the internal state of the agent that has undergone the inconsistent process as a case in which the control process of the inconsistent agent has been performed. An inconsistency processing detection program is provided.

  The invention of claim 8 is generated by providing an initial state in the virtual environment, a step of storing a control rule for controlling processing of the agent operation, and giving the state in the virtual environment to the control rule. A step of generating an internal state of the agent, and a step of storing the generated internal state of the agent, when operating the agent that is a virtual action subject in the virtual environment constructed by a simulator A method of detecting inconsistent processing, the step of storing a detection rule for detecting control processing of the agent that is regarded as inconsistent as processing of the simulator, and the detection of the processing of the agent of the simulator Monitoring the control processing of the inconsistent agent corresponding to the rule; Storing the internal state of the agent that has been subjected to the inconsistent processing as a case in which the inconsistent agent control processing has been performed when the inconsistent agent control processing corresponding to the rule is detected. An inconsistency detection method characterized by the above is provided.

According to the present invention, even in a multi-agent simulator, inconsistent processes can be detected and accumulated in a short time from the simulation results. The second embodiment, the agent inconsistency processing detection program, and the agent inconsistency processing detection method can be provided.

  Hereinafter, embodiments of the invention will be described in detail with reference to the drawings. Here, the same reference numerals are given to the same parts, and redundant description is omitted.

(First embodiment)
FIG. 1 shows a block diagram of an agent simulator and an inconsistency processing detection unit according to the first embodiment of the present invention.

  As shown in FIG. 1, the agent simulator apparatus according to the first embodiment includes a simulator control unit 1, an agent control rule set 2, an agent control unit 11, and a simulation history (log) as the main configuration of the agent simulator 100. ) Is stored.

  The simulator control unit 1 processes a plurality of agent software in a virtual environment, and controls the overall flow of simulation such as how long and in what environment when simulating the operation. In the agent control rule set 2, rule groups for performing individual processing of agents are stored. The agent is controlled by the agent control unit 11 according to a rule defined in advance in the agent control rule set 2. First, the simulator control unit 1 causes the initial state generation unit 6 to generate an initial state to be given to the agent control unit 11. The initial state is a set of environment information, agent state, and the like. These initial states may be generated from complete random numbers, or a part of the initial state created by the developer may be automatically changed by the system. Based on the given initial state and the agent control rule set 2, the agent control unit 11 processes the agent and simulates the operation. The simulator control unit 1 performs simulation for a preset number of times or time. Logs including the environment information in the simulation and the internal state of the agent are stored in the storage unit 7. The agents in the simulator are controlled by the agent control unit 11 based on the rules defined in the agent control rule set 2.

  Furthermore, in the first embodiment, the inconsistency including the inconsistency processing detection rule set 4, the monitoring unit 3, and the case storage unit 5 for automatically detecting whether or not there is inconsistent processing in the agent simulator 100. An operation detection unit 110 is provided. The inconsistent motion detection unit 200 may operate simultaneously (in parallel) with the agent simulator 100, or after the simulation operation of the agent simulator 100 ends, the internal state and environment of the simulation stored in the storage unit 7 The mismatch operation may be detected using time series data such as information.

  The monitoring unit 3 monitors the storage unit 7 storing the simulation by the agent control unit 11 or the operation history of the agent simulator 100, and when detecting an agent that matches the inconsistency processing detection rule set 4, the monitoring unit 3 The state and the state in the agent simulator 100 at the time of detection are stored in the case storage unit 5. By recording the state in the agent simulator that is accumulated slightly before the time in the simulator at the time of detection, when checking for avoidance of inconsistency processing later, the time before inconsistency processing occurs is recorded. You may be allowed to check from

  In FIG. 1, the agent control unit 11 further controls the environment display UI 220 for displaying the virtual environment information of the simulation, the state display UI 210 for displaying the internal state of the agent, the inconsistency processing selection UI 230, and the execution of the simulation. Preferably, a control UI 200 is provided. Here, UI represents a user interface. Each UI may refer to any information that is actually simulated, information stored in the case storage unit 5, or information stored in the storage unit 7.

  In the first embodiment, a BDI model having virtual personalities such as Belief, Desire, and Intention (intention) is adopted for agents as individual action subjects in the first embodiment. Belief represents information recognized by the agent, such as the agent itself and the state of the surrounding environment. Desire represents a desire (action candidate) that an agent wants to perform based on Belief. Intention represents the intention that the decision was actually made from the generated Desire.

  FIG. 2 is a flowchart for explaining processing for generating the internal state of the agent using the BDI model employed in the agent simulator according to the first embodiment.

  The simulator control unit 1 instructs the initial state generation unit 6 to generate the initial state given to the agent simulator (ST30). From the initial state generated as a result, the agent control unit 11 generates an internal state for each agent based on the agent control rule set 2. First, a Belief that the agent perceives as environmental information is generated from the initial state (ST31). Desire is generated from the set of Belief (ST32). Desire is held as a list, and one of the Desires is selected as Intention. It is determined whether or not the agent currently holds Intention (ST33), and when the agent does not hold Intention, one selected as Intention is selected from the generated Desire (ST36). A command for executing the selected intention is output, and processing is performed to operate the agent (ST37). When the agent operates, the environment information in the virtual environment changes (ST38). It is determined whether or not there is a simulation stop signal from the simulator control unit 1 (ST39). When there is no stop signal, a new Belief and Desire are generated from the changed environment information (ST31 to ST32). If the intention is already held, it is determined whether or not the existing intention is to be changed (ST34). If no change is necessary, the simulation is continued and the environmental information is changed (ST38). When it is determined that a change is necessary, a new Desire is selected as Intention (ST35). Again, a command for performing a process for executing the selected intention is output (ST37). The above operation is repeated, and when there is a simulation stop signal from the simulator control unit 1 (ST39), the simulation is stopped (ST40). The above has shown the step which the agent which concerns on 1st Embodiment produces | generates an internal state. When a plurality of agents are processed in the simulator, an internal state is generated for each agent.

  Next, a specific mode of what processing is performed in the agent simulator according to the first embodiment employing the BDI model will be described in detail with reference to FIGS. In the first embodiment, in a multi-agent simulation assuming a setting in which a human-like agent UI in a village acting as an agent draws water from a water field and extinguishes a fire that occurred in an outdoor environment as a virtual environment. is there.

  FIG. 3 is an example of the environment display UI 220 and the control UI 200. The environment display UI 220 displays the outdoor environment and the human-type agent UI 203 that embodies the agent in the simulator that operates in the environment and environmental information such as fire in accordance with the position information. Further, the control UI 200 includes a control button 201 for temporarily stopping the simulation and a time slider 202 for going back to an arbitrary time point simulated so far. A button for performing simulation at a higher speed may be added to the control button 201. Information is read from the storage unit 7 in accordance with an instruction from the developer instructed through the control UI 200. The internal state of the agent, particularly the generation rule when generating Desire from Belief, is stored in the agent control rule set 2, and is simulated according to the rule.

  FIG. 4 is a schematic diagram showing a Belief recognized by an agent from given environment information. Belief is information that an agent recognizes in given environmental information and holds it as an internal state. Rules for generating Desires corresponding to these Beliefs are stored in the agent control rule set 2.

  The position information of each object such as each agent and fire is divided into several grids, and it recognizes in which grid the agent exists and in what state, and each agent holds it as Belief. For example, as shown in FIG. 3, when the horizontal axis is the x axis and the vertical axis is the y axis, the grid position (x, y) is expressed. The agent in (3,3) has information that there are four agents in (2,3), there is a water field in (6,2), and there is a fire of strength 3 in (9,6). Recognized as Belief.

  FIG. 5 shows an example of how Desire is generated from the Belief held by the agent. In the BDI model, Desire is generated based on a control rule set in advance for a Belief that is perceived from given environment information. In addition, Desire, which is a candidate for action at that time, is generated with a rule fired by Belief at that time. In the BDI model, multiple Desires are retained for a certain period, and one of them is selected as Intention. Each agent can be appropriately processed and operated by changing which Desire is selected as Intention corresponding to various changing situations.

  For example, as shown in the figure, if an agent holds both a “Belief that the grid (x1, y1) has fire” and a “Belief that has a bucket of water”, the agent From the control rule defined in the control rule set 2, Desire is generated that “I want to extinguish the fire of the grid (x1, y1)”. Similarly, when an agent holds both a “Belief that there is a person in the grid (x2, y2)” and a “Belief that the fire strength is n or more”, the “grid (x2, y2) ) “I want to go to the person”). Also, the agent says Belief “I don't have a bucket of water”, “Belief that there is a fire in the grid (x1, y1)”, and “Belief that there is a water field in the grid (x2, y2)” When you hold it, a Desire that says “I want to draw water from the grid (x2, y2)” is generated.

  As described above, each agent generates Desire from Belief.

  Next, the agent selects one Desire from the generated Desires, and actually acts with the Desire as Intention. The agent acts to resolve the Desire selected as Intention. The selection may be performed randomly, or the Desire with the highest weight may be selected using the weight added to the above Desire. As a weighting method, for example, the weight of how much Desire the agent desires can be calculated using environmental information such as the strength of fire and the distance from the current position of the agent to the fire.

  For example, if Desire to extinguish the fire of the grid (x1, y1) shown in Fig. 5 is selected as Intention, the agent actually moves to the grid (x1, y1) and uses the water in the bucket. To put out the fire.

  Desire generation and Intention selection are performed at regular time intervals or with changes in Belief. While executing an intention, it may be set not to switch to another intention, and if a desire with a higher weight than the weight provided by the intention is generated, the intention is set to the desire. You may set so that it may switch.

  Next, a specific mode of a method for detecting inconsistent processing in the simulation according to the present invention will be described in detail with reference to the drawings. In a simulation using a rule-based BDI model, it is necessary to create an appropriate agent control rule set 2 that prevents inconsistent agent processing. By easily detecting inconsistent processing cases, it is possible to support creation of appropriate control rules.

  With reference to FIGS. 6 and 7, an example of a method for establishing a rule for determining that the agent has performed inconsistent processing in the inconsistent processing detection rule set 4 will be described in detail.

FIG. 6 is a diagram showing steps for establishing an inconsistency process detection rule in the inconsistency process detection rule set 4 of the agent. First, the simulator control unit 1 is instructed to start simulation. The initial state generation unit 6 generates an initial state to be given to the agent simulator (ST1), and the simulator is actually operated. The agent control unit 11 controls the agent according to the rules defined in the agent control rule set 2. The agent is monitored by the environment display UI 220, the control UI 200, etc., and it is monitored whether processing that is regarded as inconsistent is executed (ST2). When inconsistency processing is confirmed (ST3), the simulator is temporarily stopped (ST4). The operation control of these simulators may use the control UI 200 or the like. Then, the state display UI 210 is opened by selecting the agent that has performed the inconsistency processing by clicking (ST5). The agent is an agent having internal states such as Belief, Desire, and Intention, and the internal state of the agent is displayed on the state display UI 210. A rule that regards the set of these internal states as an inconsistent process is set and held in the inconsistent process detection rule set 4 (ST6). Select whether to continue the simulation based on the environment information at the time of detection. If you want to continue, run the simulator again (ST2). If you want to change the initial state, generate the initial state again (ST1) Make it work. If inconsistent processing is not detected as a result of the simulation or if the developer determines that inconsistent processing has been sufficiently detected (ST8), the simulation is terminated.

  FIG. 7 is a diagram showing an example (ST6) of the method for setting the inconsistency processing detection rule set 4 of the agent. The specific state confirmed by the developer as the inconsistency process in ST3 is displayed in the state display UI 210 described above. In (a) and (b), inconsistency processing is set in different cases.

  The upper diagrams in both (a) and (b) are examples of the state display UI 210 displayed on the environment display UI 220 when the specific agent regarded as an inconsistent process is pointed or clicked. The inconsistency process selection UI 230 is provided with a check box for performing selection and an inconsistency process detection rule setting button for each internal state displayed in the state display UI 210. By checking the check box corresponding to the internal state that the developer recognizes as inconsistent and pressing the inconsistency processing detection rule setting button, it is set as the inconsistency processing detection rule.

  Further, both (a) and (b) are diagrams showing the contents set in the inconsistency processing detection rule set 4 as the inconsistency processing detection rule for the set of internal states checked in the state display UI 210 in the upper diagram. .

  As shown in the status display UI210 in FIG. 7 (a), there is a combination of the internal state of the agent that the agent has a Belief that has a bucket of water and the Desire that wants to call a person. Has been generated. In this case, the developer originally assumes that when the agent holds a belief that he has a bucket of water, he / she wants to generate a Desire that he / she wants to perform fire fighting using the bucket. In this case, it can be said that inconsistent processing has been performed. The developer checks the check boxes of the inconsistency processing selection UI 230 corresponding to the inconsistent internal state, and then presses the inconsistency processing detection rule setting button to display the appearance of the agent having the combination of internal states. An inconsistency process detection rule set 4 that is regarded as an inconsistency process can be set.

  In addition, as shown in the status display UI210 (b) in Fig. 7, in the outdoor environment that is set, the operation of drawing water from the grid (10, 10) at the edge of the environment is inappropriate in terms of efficiency, so that If an agent has an Intention, it is considered an inconsistency process. After checking the check boxes of the inconsistency processing selection UI 230 corresponding to those inconsistent internal states, by pressing the inconsistency processing detection rule setting button, the agents having the checked internal states are matched by inconsistent processing. An inconsistency processing detection rule set 4 for the monitoring unit 3 to consider as operating can be set.

  When the agent holds these internal states, the agent is regarded by the monitoring unit 3 as operating by inconsistent processing.

  When setting the state of the agent specified in the check box to the inconsistency processing rule, select whether to target only the Belief related to a specific grid, or to the entire type of Belief regardless of the position of the grid. be able to. In addition, the internal state of the agent considered as inconsistent processing does not have to be a combination of all Belief-Desire-Intention. Any combination of Desire-Intention, Belief, Desire, and Intention may be used.

  When selecting the inconsistency processing rule setting button, the developer can be inquired in the dialog, and a check box for whether to include settings related to the grid position is placed next to each agent status in the status display UI 210. Alternatively, a UI may be provided in which whether or not to include the information is switched by clicking on the description portion regarding the displayed grid of the agent state.

  Further, a UI that can directly set inconsistency processing rules from a UI other than the status display UI 210 may be provided.

  By setting a combination of states that the developer does not expect among the internal states of the agent displayed in the status display UI 210, inconsistency processing that occurs in the simulator is detected, and inconsistency processing detection rules are set be able to.

  Next, steps in which the case storage unit 5 accumulates inconsistent process cases from the generated inconsistent process detection rule set 4 will be described in detail with reference to FIG.

  FIG. 8 shows steps in which the monitoring unit 3 accumulates the inconsistency processing of the agent in the case storage unit 5 based on the inconsistency processing detection rule set 4.

  The simulator control unit 1 sends an instruction to the initial state generation unit 6 to generate an initial state to be given to the agent control unit 11 (ST11). Thereafter, the simulator control unit 1 instructs the agent control unit 11 to start simulation, and simulation is started (ST12). The monitoring unit 3 monitors the history of the agent group in the simulator that is actually operating or the simulator stored in the storage unit 7, and there is an agent that matches the rule in the inconsistency processing detection rule set 4. In the case (ST13), accumulation is performed in the case storage unit 5 as an inconsistency processing example (ST14). An agent that has once caused inconsistency processing may not be a collection target of inconsistency processing in the future or for a certain time thereafter. After performing the simulation for the set time in the simulation environment set in advance by the developer (ST15), the simulator control unit 1 stops the simulator. When the initial state setting and the inconsistency processing collection from the initial state have already been performed for the initial state setting number set in advance by the developer, the process ends. Otherwise, the initial state is generated again (ST16). If simulation is performed only as specified, the simulation is stopped.

  FIG. 9 is a diagram showing an example of the inconsistency process selection UI 230 according to the first embodiment.

  The inconsistency process selection UI 230 displays a list of inconsistency processes accumulated in the case storage unit 5. Specifically, for each inconsistency process, a snapshot of the environment display UI 220 where the process occurred, the information of the agent that performed the inconsistency process, specifically information about its ID and location, and inconsistency The internal status of the agent that matches the process detection rule set 4 is displayed. For each inconsistency process, a check box is provided for setting whether or not the inconsistency process avoidance ratio is a target for calculation. The developer can use the inconsistent process selection UI 230 to select a case that actually needs to be corrected among the collected processes that are regarded as inconsistent.

  In the case of the example in Fig. 9, Agent No. 21's case is that the agent is trying to call a person with a bucket, but the location of the fire is far from the water and there is a place where there is a person going to call. Because the location is close enough, the developer thinks that the operation is not unnatural even if the fire is extinguished after calling a person, and it is not checked.

  The developer selects the inconsistent process by the inconsistent process selection UI 230, and then corrects the agent control rule set 2 to avoid the inconsistent process. In this example, if you have a bucket in the rule that derives Desire to call a person and the fire is far away from the place where the person going to call is far away, the agent control rule set 2 will not calculate that Desire To correct. A control rule editing UI (not shown) for the developer to edit the agent control rule set 2 may be provided.

  In order to confirm the inconsistency processing, a UI for shifting to the environment display UI 220 and the control UI 200 when the inconsistency processing selected by the inconsistency processing selection UI 230 occurs may be provided. Further, in order to support the correction of the rule, a UI for displaying a rule group in the agent control rule set 2 that has generated the inconsistency process selected by the inconsistency process selection UI 230 may be provided.

  In addition, a function that suppresses agents that perform inconsistency processing during simulator operation by performing processing such as eliminating agents with the internal state selected in the inconsistency processing selection UI230 or clearing the internal state of the agent. May be set in the agent control unit 11. Accordingly, it is possible to suppress agents that perform inconsistency processing without correcting the agent control rule set 2.

  Clearing the internal state of the agent means clearing the Belief, Desire, and Intention that conform to the inconsistency processing rule. When clearing Intention, the corresponding Desire is cleared at the same time. By clearing a specific Desire or Intention, a Desire with the next highest weight is selected as the Intention. Moreover, the generated Desire changes by clearing Belief.

  Further, a function of creating the inconsistency process detection rule set 4 again from the inconsistency process cases stored in the case storage unit 5 may be provided.

  In the first embodiment, the agent simulator and the inconsistent process detection unit can detect and accumulate inconsistent processes from simulation results in a short time, thereby reducing the inconsistency process detection work time of the developer. Is possible.

  The agent simulator and the program executed by the inconsistency processing detection unit according to the present embodiment have a module configuration including each function described above. The program is an installable or executable file that is recorded on a computer-readable recording medium such as a CD-ROM, floppy (R) disk, CD-R, DVD, etc. It may be provided by incorporating it in advance.

  Further, the agent simulator and the inconsistency processing detection unit may be integrally configured on the computer.

(Second Embodiment)
FIG. 10 shows a block diagram of an agent simulator and an inconsistency processing detection unit according to the second embodiment.

  FIG. 10 is a functional block diagram showing an example of a configuration for calculating the occurrence rate of inconsistency processing by the modified agent control rule set 2. In FIG. The main configuration includes a plurality of agent control rule sets 2 and a plurality of case storage units 5 for storing inconsistent processing cases corresponding to the agent control rule sets 2.

  The ratio calculation unit 300 simulates the specified amount and determines that the inconsistency processing examples have been sufficiently stored in the case storage unit 5, and then calculates the number of simulations performed and the count of inconsistency processing that has occurred. The ratio calculation unit 300 calculates the ratio at which inconsistency processing has occurred in each agent control rule set 2, and displays the calculation result for the developer using the ratio display UI 240.

  FIG. 11 is a diagram illustrating a flow for calculating a rate at which inconsistent processing occurs in a plurality of control rules.

  First, a control rule to be operated is selected (ST20). Thereafter, the initial state generation unit 6 generates the initial state of the environment information including the agent in the simulator (ST21). Thereafter, the simulation is started (ST22). When the monitoring unit 3 detects a process in the simulator of the agent that is considered to be inconsistent that matches the inconsistent process detection rule specified in the inconsistent process detection rule set 4 (ST23), the monitoring unit 3 uses the case as inconsistent process. Accumulate in storage unit 5 (ST24). Simulate in the same initial state for a preset time (ST25). Next, it is determined whether the simulation has been performed by generating the initial state for the set number of times. If the number of times is not full, the initial state is generated again (ST27). After determining that the simulation has been performed only in advance, it is determined whether there is a control rule that has not been simulated yet. If there is another control rule that has not been simulated yet, the simulation and inconsistency detection / accumulation operations are repeated using the other control rule (ST28). After determining that the simulation has been performed for all the control rules to be simulated, the number of inconsistency processing cases accumulated in ST24 is calculated for each control rule, and the proportion of inconsistency processing is calculated (ST29).

  By repeating the correction of the agent control rule and the confirmation of the inconsistency processing avoidance ratio, it is possible to evaluate the occurrence ratio of the inconsistent operation accompanying the change of the control rule. Based on the evaluation results, the developer can create an agent control rule set 2 with a low probability of causing inconsistency processing.

  In ST24, the environment information in the virtual environment in the state detected as inconsistent processing and the internal state of the agent detected as inconsistent processing may be stored. Further, only the inconsistency processing performed may be counted and stored in the case storage unit 5, and the ratio calculation unit 300 may calculate the inconsistency processing occurrence ratio based on the number of inconsistency processes. . The agent simulator and the inconsistency processing detection unit according to the second embodiment have a mechanism for evaluating a plurality of created agent control rule sets 2. As described above, in a rule-based system, it is difficult to predict how the rule will fire, and modifying the control rule may increase the probability that other inconsistency processing will occur. Therefore, in order to evaluate the modified rule, a simulation is performed based on a plurality of control rules, and the validity of the control rule is evaluated based on the result of the inconsistency processing detection rule set. At that time, evaluation is performed by calculating the probability of occurrence of inconsistency processing for each of the plurality of created agent control rule sets 2.

Block diagram of the first embodiment of the present invention The flowchart which shows the process which produces | generates the internal state of an agent by a BDI model. The figure which showed an example of simulator environment display and control UI. The schematic diagram showing the state of the environment which the agent recognizes and holds as Belief. The figure which showed an example of agent state display UI. The flowchart which shows the operation | movement which prescribes | regulates an inconsistency process detection rule. The figure which showed an example of the inconsistency process detection rule set. The flowchart which shows the operation | movement which the case memory | storage part accumulate | stores an inconsistency process case. The figure which showed an example of inconsistency process selection UI. The block diagram of the 2nd Embodiment of this invention. The flowchart which shows the operation | movement which calculates the inconsistency process generation | occurrence | production ratio.

Explanation of symbols

1 Simulator control unit
2 Agent control rule set
3 Monitoring unit
4 Inconsistency detection rule set
5 Case memory
6 Initial state generator
7 Memory unit
11 Agent control section
100 agent simulator
110 Inconsistency detection detector
200 Control UI
201 Control button
202 Time Slider
203 Agent UI
210 Status display UI
220 Environment UI
230 Inconsistency selection UI
240 percentage display UI
300 Ratio calculation part

Claims (8)

An initial state generator for giving an initial state in the virtual environment;
A first storage unit storing a control rule for controlling processing of an agent operation;
An internal state generation unit for generating an internal state of the agent generated by giving the state in the virtual environment to the control rule;
As a process of operating an agent that is a virtual action subject in a virtual environment constructed by a simulator device including a first storage unit that stores the internal state of the agent generated by the internal state generation unit. An inconsistency processing detection device for detecting a consistent processing,
A second storage unit that stores a detection rule for detecting the control process of the agent that is regarded as inconsistent as the process of the simulator device;
A monitoring unit that monitors the control processing of the inconsistent agent corresponding to the detection rule among the processing of the agent of the simulator device;
When the monitoring unit detects the inconsistent process, the monitoring unit includes a second storage unit that stores an internal state of the agent that has been subjected to the inconsistent process as a case in which an inconsistent agent control process has been performed. Inconsistency processing detection device characterized by things.
2. The mismatch processing detection apparatus according to claim 1, wherein the agent holds belief, desire, and intention as a virtual personality in the first storage unit as the internal state.
The detection rule is defined based on the internal state, and the monitoring unit regards the inconsistent process when the agent holds an internal state corresponding to the detection rule. 3. The inconsistency processing detection device according to 2.
First display means for displaying the internal state of the agent;
A first selection means for selecting the internal state as one of detection rules when the internal state is a combination of states not assumed by the developer;
2. The inconsistency processing detection apparatus according to claim 1, further comprising means for storing the selected detection rule in the second rule set.
Storing the processing regarded as the inconsistency stored in the second storage unit;
The inconsistency processing detection apparatus according to claim 1, further comprising a calculation unit that calculates a probability that the inconsistency processing occurs.
6. The inconsistency processing detection apparatus according to claim 5, further comprising second display means for displaying a probability of occurrence of the inconsistency processing calculated by the calculation unit.
The ability to generate an initial state in the virtual environment;
The ability to store control rules that control the processing of agent actions;
A function of generating an internal state of the agent by giving a state in the virtual environment to the control rule;
A program for detecting inconsistent processing when operating the agent that is a virtual action subject in the virtual environment constructed by a simulator having a function of storing the generated internal state of the agent. And
A function for storing a detection rule for detecting the control process of the agent, which is regarded as inconsistent as the process of the simulator;
A function of monitoring inconsistent agent control processing corresponding to the detection rule in the agent processing of the simulator device;
A function to store the internal state of the agent that has been subjected to the inconsistent process as a case in which the inconsistent agent control process has been performed when the inconsistent agent control process corresponding to the detection rule is detected; An inconsistency processing detection program characterized by being realized by a computer.
Providing an initial state in the virtual environment;
Steps stored by control rules that control the processing of agent actions;
Generating an internal state of the agent generated by giving a state in the virtual environment to the control rule;
A method for detecting inconsistent processing when operating an agent that is a virtual action subject in the virtual environment constructed by a simulator comprising a step of storing the generated internal state of the agent, ,
Storing a detection rule for detecting the control process of the agent, which is regarded as inconsistent as the process of the simulator;
Monitoring the control processing of the inconsistent agent corresponding to the detection rule in the processing of the agent of the simulator;
Storing the inconsistent agent internal processing as a case in which the inconsistent agent control processing is performed when the inconsistent agent control processing corresponding to the detection rule is detected; An inconsistency processing detection method characterized by comprising.

Images