JP2011258166A - Autonomously operating agent system that makes self evolution by learning function - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a system that autonomously operates by learning user's daily activities and provides a function more suitable to the user's activities by self evolution.SOLUTION: An agent 34 acquires a moving destination spot from information existing in a spot manager and uses a function of the spot. The agent analyses an accumulation result of the user's daily activities by trend analysis, error analysis, digitization of information, analysis of repetitive activities, and the like, modifies these scripts as needed by a specific logic, adapts the function to an actual state, and constructs a predictive logic. The agent can receive an event as needed by registering a callback of an event notification to an event manager 35-2. The agent operates by using this event as a trigger, and performs an event driven autonomous operation.

Description

  The present invention relates to an agent system that autonomously operates on the basis of event-driven processing and the function of a self-evolutionary object equipped with a learning function and an interpreter language, and self-evolves with the learning function.

  Japanese Patent Application Laid-Open No. 10-307804 discloses means by which a character evolves to create a new action from communication with a user and communication between characters. By using such a technique, it is possible to provide an autonomous evolution type character action generation method and apparatus that allows a character to generate a new action evolutionarily through communication with a user.

Japanese Patent Laid-Open No. 10-307804

Problems to be solved by the invention

  However, the conventional example is a system in which the character changes only when the user takes an action, and there is a problem that the character itself cannot start an action.

  Furthermore, although the action is generated by controlling the action of the character by the finite automan, if the finite automan is used for the control, it is necessary to describe a state machine (state machine) in advance, which is necessary in the present invention. There is a problem that self-modification cannot be realized.

  In addition, the character evolves due to communication with the user, but in this case, there is a problem that the character cannot evolve due to the relevance to the location information, time, and external factors by the push from the server.

  The present invention has been made to solve the above-described problems, and an object thereof is to provide an agent system that operates autonomously and is self-evolving by a learning function, which is convenient for the user.

Means for solving the problem

  In order to achieve the above object, according to one aspect of the present invention, an agent system that autonomously operates and self-evolves by a learning function receives an event due to an external factor and learns a user's behavior and function In addition, it has a function of predicting user's behavior by a function of self-evolution by modifying the interpreter language that determines its own action, and the data for the action is held while being synchronized between the server and the client. .

  The agent system that preferably operates autonomously can receive a callback due to an external factor by registering a callback method in the event manager, thereby realizing event-driven.

  Preferably, the agent system that operates autonomously performs the behavior analysis by counting and accumulating the user's repeated behavior, and sums and accumulates the probability that the self-evolved function matches the user's actual behavior. The stage of performing behavioral analysis, the amount of error that the self-evolved function differs from the actual behavior of the user, the stage of changing the error threshold and changing the probability of matching, and the specific function It comprises a step of performing behavioral analysis by counting and accumulating the frequency of execution, and a step of performing behavioral analysis by counting and accumulating those of a specific function group in the same field.

  In an agent system that operates autonomously and self-evolves with a learning function, synchronization between information held in the terminal and information held in the server is performed implicitly in the background, and the same data from any terminal The agent can be operated using

  An agent system that operates autonomously and self-evolves through a learning function has a means for an agent object to acquire a method that can be executed from a spot with a specific function that comes into contact with an agent that operates autonomously as a script. The agent issues a query to the spot manager and obtains spot information to be moved next.

  Note that registering callbacks for a plurality of events has the advantage of being able to act more autonomously.

  In addition, an agent system that operates autonomously and self-evolves with a learning function provides a class library for developing spots to generate unique spots, and also for developing agents to generate unique agents. A class library is provided.

  Further, an agent system that operates autonomously and self-evolves by a learning function includes an interpreter compiler for executing a script.

It is a block diagram of the terminal carrying the agent system in one of the embodiments of the present invention. It is a block diagram of an agent system in one embodiment of the present invention. It is a flowchart which shows the operation | movement of "autonomous" in this Embodiment. It is a flowchart of the process of step 403 of FIG. It is a flowchart which shows the operation | movement of "accumulation" in this Embodiment. It is a flowchart which shows the operation | movement of "analysis" in this Embodiment. FIG. 10 is a sequence diagram showing an operation of “self-modification” in the present embodiment. It is a use case figure which shows the operation | movement in this Embodiment.

  FIG. 1 is a block diagram of a terminal equipped with an agent system according to one embodiment of the present invention.

  Referring to the figure, the agent system includes event managers 35-1 and 35-2 for managing events, function managers 36-1 and 36-2 for managing spot functions, and agents that are main operating objects. 34, a database 33 that stores information necessary for operation, a script engine 41 that analyzes and executes an interpreter language for autonomous operation, and a spot manager 42 that stores function indexes of a plurality of spots It is composed of

  The event managers 35-1 and 35-2 have a function of notifying the agent of location information of a user via a location information system (GPS) mounted on the terminal, a timer, and further information notified from the server, The agent registers a callback method here. The function managers 36-1 and 36-2 are portions where functions in the spot are registered. The database 33 stores information necessary for the operation of the agent system, and has functions accessible from the function managers 36-1 and 36-2 and the agent 34.

  The spot manager 42 has a function of managing spots and returning a function index and a spot type held by each spot when a query is received from an agent. The script engine 41 is an interpreter analysis execution engine for analyzing and executing a function script held by the function manager.

  Next, the configuration of the terminal equipped with the agent and the server will be described.

  FIG. 2 is a block diagram of an agent system in one embodiment of the present invention.

  Referring to the figure, the agent system includes a database 201, a server 202 for synchronizing terminal data and data on the database, a back-end system 203, an Internet network 301, and a terminal 101. .

  The server 202 is connected to the Internet 301 via a route control device or a telephone line. The terminal 101 includes a reaction unit having an input / output device, an audiovisual notification function, a position information acquisition notification function, and a timer function, and includes a wide area network wireless communication function such as packet communication, line communication, PHS data communication, and wireless LAN. ing.

  Next, an event-driven and autonomously operating process in the agent system according to the present embodiment will be described in detail with reference to the flowchart shown in FIG.

  Referring to FIG. 3, first, as indicated by F <b> 10, terminal 101 is kept waiting for callback reception.

  When an event occurs in the terminal 101 and a callback is notified, the terminal 101 detects which event has occurred and activates a specific event handler registered in advance. Each event handler wants to describe a specific process, and the process is executed accordingly. FIG. 3 describes typical events. By executing processing with an event as a trigger, it is possible to operate autonomously without waiting for a user instruction.

  FIG. 4 is a flowchart describing in detail the date / time callback reception process in the flowchart of FIG. When a specific event, which is the flowchart shown in FIG. 3, has occurred, it is checked whether another event has occurred and the content is saved. Notify all events that finally occurred.

  FIG. 5 is a flowchart showing the accumulation operation. A specific method registered in advance using an event as a trigger is executed, and accumulation of frequency, deflection, probability, repetition, error, and the like is performed based on information from the device and a user's reaction. A learning function is implemented by this accumulation.

  FIG. 6 is a flowchart showing the analysis operation. Trend analysis (normal distribution), prediction error analysis, set value validity analysis, iterative behavior analysis, regression analysis, variation analysis (variation coefficient), correlation analysis, etc. Analyze using the analytical method. This analysis is self-modifying logic.

  FIG. 7 is a flowchart showing a self-modifying operation. As indicated by F82, in the case of the dominant modification, each modification can be preferentially modified. If it is an inferior modification, the fact that it is inferior is stored, and for each modification, only the modification in the case of inferiority is kept. Such modifications enable self-evolution.

Further, by changing a script that determines its own operation, it is possible to perform self-modification while continuing to operate.

  FIG. 8 is a use case diagram showing the operation. The processes from FIG. 3 to FIG. 7 are shown as use cases. The event that occurred in the terminal indicated by F104 calls the event handler of F109 and is notified (F107) to the user (F101) by the method described therein, and the user executes it based on it (F105), The selection and execution results are accumulated (F106) and analyzed (F108), and stored in the database of the terminal indicated by F102.

  As described above, according to the present embodiment, it is possible to perform an accurate behavior prediction without taking a complicated procedure by combining an autonomous operation and a function of self-evolution by a learning function. There is an excellent effect.

  The embodiment disclosed this time should be considered as illustrative in all points and not restrictive. The scope of the present invention is defined by the terms of the claims, rather than the description above, and is intended to include any modifications within the scope and meaning equivalent to the terms of the claims.

30 Application 32-1, 32-2 Application-specific function storage library in terminal 33 Data storage unit in terminal 34 Autonomous driving control unit 35-1, 35-2 in terminal Event management unit 36-1, 36-2 in terminal Application specific function (script) unit 40 terminal agent system platform unit 41 terminal script engine 42 terminal application specific function management unit 101 terminal 201 database on server 202 server 203 for communicating with terminal backend system 301 Internet network F10 to F25 Step F30 to F40 Step F50 to F65 Step F70 to F77 Step F81 to F93 Step F101 to F109 Step.

Claims (10)

  A method for an agent object to acquire a method that can be executed from a spot with a specific function contacted by an agent having an autonomous function as a script, and a spot that the agent issues a query to the spot manager and then moves An agent system that operates autonomously and self-evolves by a learning function, characterized by comprising means for acquiring information.   Most of the information necessary for executing the agent is managed by the server, only a part of the necessary information is stored in the terminal, and backed up on the terminal when communication with the server is online. An agent system that autonomously operates and self-evolves by a learning function, comprising means for automatically synchronizing in the ground and means that can be used from any terminal regardless of online or offline.   Means for registering a method for receiving a change in position information from the GPS function of the terminal and notifying the agent, means for registering a method for notifying the agent together with the timer type when the time set by the timer comes, and a server An agent system that autonomously operates and self-evolves by a learning function, comprising means for registering a method for notifying an agent of a specific event from the system by a push function.   2. The agent system according to claim 1, wherein the agent system is autonomously operated and self-evolves by a learning function. Self-evolving agent system.   An agent system that operates autonomously and self-evolves with a learning function, characterized by generating a prediction function by accumulating the probability that a specific event will occur.   Means for generating a prediction function by accumulating that a particular event is repeated, means for generating a prediction function by accumulating the probability that a particular method is called, and results in the execution of a particular method A means for adapting to the method caller without changing the threshold itself when the probability that the threshold exceeds the predetermined threshold exceeds a predetermined rate, and adapting to the method caller, thereby evolving itself An agent system that operates autonomously and is self-evolving with a learning function.   7. The agent system that operates autonomously and self-evolves by a learning function according to claim 6, comprising means for changing the script itself acquired from the spot and making it self-evolve, and an interpreter compiler for executing the script, Agent system that operates autonomously and self-evolves with learning functions.   In the agent system which operates autonomously according to claim 1 and self-evolves by a learning function, means for providing a class library for developing a spot to generate a unique spot, and for generating a unique agent An agent system that autonomously operates and self-evolves by a learning function, comprising means for providing a class library for developing an agent and an interpreter compiler for executing a script.   10. Self-evolution by autonomously operating and learning function according to any one of claims 1 to 9, characterized by providing food management and menu recommendation services as applications running on systems based on agent systems Agent system. 10. The agent system according to claim 9, wherein the agent system autonomously operates and self-evolves by a learning function, the preference is quantified from selection establishment, and analysis / self-evolution is performed based on the numerical value.

Images