This paper discusses a framework and the algorithms for a reconfigurable intelligent system, which is capable of "knowledge amplification". The proposed intelligent system consists of many kinds of knowledge sources, which have simple levels of data and intelligence together with some blackboards which dynamically organize a hierarchical structure in order to optimize its own performance index. The system has two strategies for reconfiguration of the system structure: one is "entropy maximum strategy" the other is " entropy minimum strategy". The entropy maximum strategy is used when many kinds of tasks are given to the system. The entropy minimum strategy is applied in cases were the system deals with a few tasks only. This paper addresses the fundamental configuration and algorithms which include these two strategies for reconfiguration. The efficiency of the proposed system is shown by some simulations.>

The authors have studied the cellular robotic system (CEBOT) with the aim of realizing a self-organizing robotic system. It can be regarded that CEBOT is one of the multi-agents robotic systems studied by many researchers. It is generally considered that one of the features of the multi-agents robot is that the system can carry out many kinds of tasks under various environments due to coordination among many agents. It is assumed that a group of the agents can get higher performance than an ability of single agent. In this paper, the relation between the resource amount, which implies both hardware and software installed into the system, and system performance is discussed. At first, a new concept called function amplification is proposed and defined, the control law of each agent to achieve the function amplification is secondarily discussed. In the last part of this paper, some simulations are reported and discussed to demonstrate the relation between resource amount and system performance.

The authors propose a genetic knowledge production algorithm (GKPA) for CEBOT, in order to realize a self-organizing and self-evolutionary knowledge system. They propose a new kind of distributed intelligence system which is based on the concept of CEBOT employing the GKPA. This proposed intelligence system has capabilities of learning, reasoning, and self-organizing. This system consists of various kinds of intelligence units called knowledge-cells. As an example of the applications of the self-organizing intelligence, a robotic manipulator capable of learning and reasoning has been made from the CEBOT concept by installing the intelligence system with GKPA in the manipulator. The knowledge acquisition ability of the manipulator is demonstrated by experimental results.>

Cellular Robotic SystemCEB0T has been studied to realize a dynamically r econfigurable r obotic system, which can c arry out m any kind of tasks under various environments by the authors. CEBOT consists of a lot of units called cell which has a simple function. To deal with a complex task which can not be carried out by a cell, CEBOT generates various module ,which is composed of some cells. A much complex 5ystem composed of some modules is called structure, which can carry out very complicated tasks. As above mentioned concepts, CEBOT is very similar to a living creature, because tissues and organs are formed by a lot of cells. CEBOT h as to not only create new modules and structures, but also kill some modules and structures, because the number of cells is limited. Therefore it is necessary to study a system which can create and kill new group composed of many units of the system. In this paper, We propose a strategy of self-organization. We call the strategy an Genetic Cell Production Algorithm (GCPA). The GCPA is based on the Genetic Algorithms, which are search algorithms based on the mechanics of natural se1ectiol.r and n atural genetics. We also propose a estimation method for the distributed system, which can be described by relations among subsystems. To simulate the self-organization based on the GCPA, we designed a distributed s ystem, which has a hierarchical cell-structure based on the concept of CEBOT. By using the model of a distributed system, some simulational results and analysis are reported in this paper, to verify the efficiency of the GCPA.

The dynamical self knowledge allocation method is presented for a distributed intelligent system. The authors apply the method to the cellular robotic system CEBOT, which is one of the distributed systems. They propose a new matrix notation, the knowledge-based structure matrix, to express the communication link and states of the knowledge allocation. They also propose a matrix notation for describing a given task, the task matrix. The efficiency of the dynamical self knowledge allocation method reduces the total system communication information volume.>

'Cellular robotic system (CEBOT)' consists of many units called 'cells'. CEBOT is required to be able to carry out many different tasks under various surroundings. The authors propose a new kind of distributed intelligent system which is based on the concept of CEBOT. This intelligent system proposed has the abilities of learning, reasoning, and self-organizing. Similar to the hardware structure of CEBOT, this system consists of various kinds of 'intelligence' units called 'knowledge-cells'. By using the knowledge-cells and the proposed algorithms, the system which can generate task procedures is realized. To estimate the efficiency of this system, a search space based on the theory of probability is described.>

The authors propose an optimal structure decision method which can determine cell type, arrangement, degree of freedom, and link length. It can be used for fixed-base and mobile-base manipulators, so that universal/modular manipulators can be made. Simulation results show the differences in evaluation for different candidates and tasks and also demonstrate that the structure decision method works efficiently. With respect to the software structure, optimal knowledge allocation is one of the most important issues for CEBOT. The authors propose a method of optimal knowledge allocation based on the amount of communication among cells. Communication among cells for the control proposed is described.>

A dynamically reconfigurable robotic system (DRRS) is one that can reconfigure itself to an optimal structure, depending on the purpose and environment. To realize this concept, the authors propose CEBOT (cell structured robot), which is a distributed robotic system consisting of separable autonomous units. These functional cells are able to communicate with each other and to approach, connect, and separate automatically. If single cells of CEBOT are damaged, they can be repaired or replaced automatically. Since CEBOT is capable of adapting itself to changing environments, it is a very flexible system applicable in space, factory, and hostile environments. The authors propose an optimal structure decision method that can determine cell type, arrangement, degree of freedom, and link length. It is applicable to fixed-base and mobile-base manipulators. A structure evaluation function, which is the sum of parameters relating to the number of work points, required positioning accuracy, torque, distance between given work points, and cell cost, is presented. Simulation results are given.>