The Agent Layer Concept and Ubiquitous Concept Databases
Keiji Mitsubuchi
Nihon Web Concepts, Ltd.1
1-15-19 Mejirodai, Bunkyô-ku, Tokyo, 112-0015
January 26, 2003
Abstract
This paper proposes the “agent layer concept” as the infrastructure for constructing a ubiquitous information society. The agent layer concept separates information transfer in today’s society into 5 layers from the standpoint of agent functions. It employs a ubiquitous concept database at its core to provide the infrastructure necessary to an information society. Research on the feasibility of implementing such a concept will initially focus on limited development related to client platforms and the forms in which information will be exchanged.
1. Introduction
The accelerating pace of computerization in our modern society has given rise to a number of problems, including confusion resulting from a glut of information and violation of individuals’ privacy. Recently, notions such as broadband and ubiquitous computing2 have permeated throughout the whole of society, and the above-mentioned problems can only be expected to grow more serious if technological development continues to proceed in a piecemeal fashion. There is thus an urgent need to design an infrastructure for building a ubiquitous information society.
To this end, I propose the agent layer concept as a means of solving these problems and providing necessary information in a ubiquitous manner. This concept, modeled on the information transfer mechanism of the human nervous system, classifies the venues of information transfer in present-day society into 5 layers from the standpoint of mobile agent functions. I further propose the “ubiquitous concept database ”3 as the database positioned in the top layer, by which decentralized control is exercised over agents and data. In this interesting and novel concept, the agents absorb differences in conditions existing at the various layers, enabling users to obtain and transmit the necessary information in a ubiquitous manner (without consideration of the situation in which the other party exists).
The content of this paper is as follows: In the next chapter, Chapter 2, I introduce the agent layer structure, which classifies the flow of information in society (in a form including user and the environment). I then discuss the relationship between the individual layers and agents, as well as the ubiquitous concept database that forms the core of this concept. The content of Chapter 2 presents the main theme of this paper, and here I make my conceptual proposals. In Chapter 3, I touch upon limited development and areas of application at an intermediate level as the first stage of research toward implementation of the agent layer concept. Lastly, in Chapter 4, I sum up the proposals made in this paper.
2. The Agent Layer Concept
The agent layer concept that I propose has as its object the construction of an information society infrastructure that would enable users to obtain and transmit necessary information irrespective of place, time, or devices. Here, the transmission and processing of information are performed by agent functions, with the agents controlled and optimized by a new type of database. Let us first consider the necessary infrastructure for implementing such a paradigm. I have attempted to represent the various functions and roles in Figure 1, dividing the flow of information in society into 5 layers from the standpoint of agents.
In accordance with Fig. 1, the layer structure of the flow of information can be outlined from the bottom up as follows. First, Layer 0 (the Environment Layer), indicates the natural world and society at large, including the user. In other words, it encompasses all things (not limited to human beings) that provide and receive information in various forms. If we liken this system to the information transmission mechanism of the human nervous system, Layer 0 corresponds to the stimulus, or Event. Layer 1, the Device Layer, indicates devices that mediate between the environment and the world of the electronic brain. In our analogy with the human nervous system, these correspond to the five senses (sight, hearing, taste, smell, and touch). Layer 2, the Personal Layer, indicates the server that performs processing when an instantaneous response is required. In human terms, this corresponds to the muscular reflex action effected through the motor and sensory nerves. Layer 3, the Agent Layer4, is a center clustering the subordinate server groups, and it is responsible for creating and controlling the agents. In our human analogy, this layer performs functions of the spinal cord and cerebellum. Finally, we have Layer 4, the Database Layer, whose objective is the management and storage of a tremendous volume of data. This layer’s role is akin to that of the human cerebrum. Next, I shall describe the workings of each layer in more detail.
2.1 Device Layer
The device layer occupies the bottommost stratum in the layer structure, where it comes in contact with the natural world and human society to effect the exchange of information. General I/O devices (keyboard, mouse, monitor, etc.) as well as sensors are included here (See Fig. 2).
The device layer transmits collected information from the environment to the personal layer above. Devices in the agent layer concept differ from presently existing devices in that they themselves function in the place of agents. This makes the following two things possible.
(1) Even if the standards for individual devices vary from maker to maker, the user can use the devices without any special setup. This is because information on device functions and other characteristics is transmitted from maker sites and databases to the personal layer through the agent layer above, and here the individual differences among the devices are absorbed.
(2) When the individual devices are occupied by agents with self-diagnosis, function change, or other procedures, it automatically becomes possible to utilize such functions from the personal layer through communication with the layer above. For instance, in the case of operating a vehicle, information on the condition of the engine and brake control can generate recommendations to the driver on when to replace parts, etc., making it possible to provide a safer driving environment.
2.2 Personal Layer
The personal layer is positioned above the device layer, and it consists of personal centers linking PCs, PDAs, car navigation systems, etc. These personal centers possess the functions requisite for instantaneous response in the control of input/output from devices, as well as in local data storage and processing of information with a high frequency of use. In addition, it sends requests for new agents to the agent layer above and performs backup and duplication of data in storage to the layer above. For example, let us consider a case in which a personal center is constructed linking monitors, sensors, etc., within an automobile (See Fig. 3).
Through the transfer of information between the speedometer/sensors and the car navigation system during operation of the vehicle, it is possible to provide the driver with information on the neighborhood and maps to the destination in an efficient manner. Private information collected from the various on-board devices is also stored here, and service fine-tuned to meet the needs of the driver can also be provided. When Bluetooth is used to exchange information with personal centers in other cars, it also becomes possible to provide such services as collision-prevention systems and systems for identifying areas of traffic congestion.
2.3 Agent Layer
The agent layer is positioned above the personal layer, forming the nucleus for all services in the agent layer concept. This layer is comprised of agent centers clustering a number of servers creating and controlling agents, as well as other servers responsible for managing data (See Fig. 4).
In this layer, the agents check the integrity of the data within the personal centers in the layer below and create backups for it. The establishment of agent centers linked by high-speed fiber in various parts of the world would allow for global data sharing. Connection of agent centers with the Internet and telecommunications and wireless networks would also enable it to function as a gateway to these networks.
2.4 Database Layer
The database layer holds the topmost position in the layer structure, and it is responsible for data management and agent optimization (See Fig. 5).
Private/Group/Public attributes are added to all of the data managed in the database layer, and all of the agents’ authority corresponds to these individual attributes. This managing system speeds up the service and enhances security and tuning to individual users. Moreover, the agents are continually optimized by the knowledge stored in the database layer. A number of servers possessing data management and agent optimization functions of this kind cluster together to form one database center. A newly conceived ubiquitous concept database is employed as the database management system at the heart of the center to maximize the advantages of information transmission by agents in the layer structure.
I shall conclude this chapter by describing the ubiquitous concept database in greater detail. In this database management system, the users’ history of data use and solutions offered by the system are selected in the form of “concepts” and accumulated in the database as knowledge. A system of management by ubiquitous concept database is one in which such knowledge and various kinds of data are managed in a usable form that can accommodate ubiquitous computing.
The method used to manage knowledge and data in this managing system differs from that used in the existing relational databases5 and is based on a new concept. Each piece of data is recorded as one element of a sub-space that characterizes it, and the log at the time of entry and the relations to other elements are updated in dictionary form. Therefore, each time the user adds various data to the database, this system semiautomatically creates relations and performs data segmentation on the basis of the user’s data-use history and dialogue with the user. Let us consider the example of entering Mr. A’s telephone number, ***-***-****. In the management method employed in existing relational databases, these data are recorded as additions to a table relating “name” with “telephone number.” In this system, however, “Mr. A” is recorded as an element of the sub-space characterized by “name-like things,” and “***-***-****” is recorded as an element of the sub-space characterized by “phone number-like things”; then the relationship between the two elements is updated as a dictionary entry. Because the system uses a concept-based managing method, there is no need to design tables defining relations; it is sufficient to take the set union of the sub-spaces even when integrating databases.
Another noteworthy characteristic of this database managing system is the way in which message information is consolidated. All pieces of message information, whether e-mails, Web pages, telegrams, telephone voice mail messages, etc., are brought together as elements of a sub-space characterized by “message-like things,” independently of the environment at the time of sending or receiving. Thus, when sending a message, one need not be conscious of whether the recipient will read the mail at the PC in his or her own home, or use a PC at work, or listen to the message as voice mail on a cellular telephone. The party receiving the message will simply go through an agent to access the sub-space where the message addressed to him is stored, and then read the message. This function not only saves time and labor for the service provider, it also prevents redundancy in service and reduces excess information volume.
3. Limited Development and Areas of
Application
The agent layer concept that I propose would function as a large-scale infrastructure for building a ubiquitous information society, and tremendous investments of time and resources would be required to implement this concept. Moreover, from the standpoint of the convergence of data sharing and concept-based recommendation functions, it would not make sense to create such an infrastructure unless it could be established as a de facto standard. Accordingly, system development in as open a form as possible would be desirable, and to this end, wide-ranging cooperation on the part of industries and countries would be indispensable.
As the first stage of research toward implementation of this concept, I envision limited development: the only kind of data handled would be text strings, and devices would be limited to the keyboard, mouse, monitor, and speakers. Personal centers would consist of notebook-type computers (Win2000), PDAs, and cell phones (iApps), and on that foundation, a platform for use in data storage and by data management agents would be developed. In addition, the agent centers and database centers would be constructed on the same Linux server for ubiquitous data exchange between different personal centers, issuing of agents, and development of gateways from agent centers to the Internet and mobile information networks. The system would target multiple users. Clustering of data in the database would be done offline using a statistical technique6, and the results would be reflected in personal centers.
In the limited development described above, application as small-scale groupware would be possible. At present, we are considering development as an information provider service client linked with the Sankei Shimbun’s news delivery service7. I would like to see the client software itself utilized broadly free of charge to users and applied toward future development.
4. Conclusions
In this paper, I have proposed the agent layer concept as an infrastructure for providing needed information in a ubiquitous manner. Under this concept, the flow of information in present-day society is divided into 5 layers. Transfer and processing of information between layers is performed by mobile agent functions, and these agents are managed and optimized by a new type of database. This concept has the following 3 characteristics not found in any existing systems:
1. “Ubiquitous Concept Database” Based on a New Concept
This new database managing method facilitates concept-based searches and makes it possible to obtain the necessary information quickly. It also eliminates the existing difficulties with database integration work.
2. Data Sharing
Private/Group/Public attributes are added to all data for management purposes, and agents function in accordance with each attribute, allowing for tuning of service to individual users, speeding up of service, and strengthening of security.
3. Message Function
Because agents absorb differences in environment within each layer, there is no need for the party sending the message or the party receiving it to be personally aware of each other’s environment. This function saves time and effort on the sender’s side and also makes it possible to prevent issuing of redundant messages and reduce excess information volume.
With these new features, the agent layer concept can provide a paradigm for solving many problems in our present information society. For initial inspection as the first step toward implementing this concept, we are planning to construct a ubiquitous information provider environment targeting a small number of users on a single server.
About
the Author
Keiji Mitsubuchi
Chief Executive Officer
Nihon Web Concepts, Ltd.8
As a student, Keiji Mitsubuchi researched image recognition and was interested in AI and clustering methods for criterion function spaces. After graduation, he gained experience in inference engine / learning system R&D at OMRON’s Santa Clara Research Lab. He presented at the International Fuzzy Association on fuzzy automatic rule generation systems and holds U.S. patents in related research. In recent years, Mitsubuchi acquired experience in proposing and developing language recognition system algorithms to reject obscene expression or slander on the Internet, as well as in research and development of recognition systems for pornographic images. These experiences yielded fundamental technology essential to the segmentation and relation of data in the project proposed in this paper.
Mitsubuchi also possesses a great deal of practical business experience. His involvement in the development of recommendation systems through his business relationship with a major telecommunications company, as well as his work designing, developing, and operating a news and e-mail delivery system for the Sankei Shimbun9, has provided him with a wealth of experience related to practical databases of public benefit. The experience gained through the production, administration, and operation of formal content for cellular phones, which is the core business of Nihon Web Concepts, will also provide technologies indispensable to ubiquitous information providing.
Notes
2. “Ubiquitous” means “existing everywhere” in Latin. “Ubiquitous computing,” an expression coined by Mark Weiser (1988), refers to a computing environment in which a number of computers are embedded in all the different venues of work and everyday life so that users can employ them without even being conscious of doing so.
3. Here, the concept database signifies a database that accumulates, stores, and manages data and knowledge in the form of concepts. It can therefore handle not only “and/or” searches, but also lends itself to concept-based search functions. For details, see Chapter 2.
4. The agent layer described here ranks as the nucleus of information transmission by the agents, but agents also exist and function in other layers as well (with the exception of the environment).
5. In a relational database, the data is managed in the form of a table defining “relations” with other data. Initial design of such tables complicates the construction of databases and makes the work of integration more difficult.
6. Because the number of users in this limited development would be small, we would not use learning algorithms (fuzzy neural nets, etc.) requiring very large numbers of examples for convergence. If this initial stage of development is fruitful, we plan to gradually switch over to a combined form incorporating learning algorithms.
Figures
Fig. 1. Schematic diagram of the agent layer structure. The flow of information in society is classified hierarchically. There are 5 strata or layers (including the environment), and transfer of information within the same layer or between different layers is performed by agents.
Fig. 2. Example of a device layer. Various devices, including I/O devices, fulfill mediatory roles, acting as a bridge between the environment and the world of the electronic brain.
Fig. 3. Example of a personal layer. Personal centers are constructed in each car by connecting monitors, sensors, etc., within the car. It is also possible for different personal centers to exchange information using cell phones or Bluetooth.
Fig. 4. Schematic diagram of the agent layer. Here, servers creating and controlling agents cluster together with other servers managing data to form one agent center. An agent center is connected with other agent centers as well as with the Internet.
Fig. 5. Schematic diagram of the database layer. Database centers are formed by clustering numbers of servers managing various kinds of data, ranging from common data to data attuned to individual users.