ROBOTICS

[This page is CSS2 enabled. Your browser might not fully support it]

[http://www.ee.oulu.fi/research/robotics/senmp/]

REDI - Remote Distributed Intelligence

$RCSfile: index.html,v $ $Revision: 1.2 $ $Date: 2005/12/01 06:58:12 $

ABSTRACT

Distributed intelligence is concerned with the design of a multi-agent system that achieves a global objective while the system itself is composed of agents which are inherently distributed in space, time, or functionality. For example, a team of robots performing a concerted mission describes a distributed intelligent system. The application possibilities of multi-agent systems are appealing as concerted missions can be performed without often complex, non-robust, non-scalable, and time-consuming centralized control. Rather, by utilizing a fault tolerant multi-agent distributed intelligent system in which the global behavior of the system emerges from local interactions of agents with their environment and with each other, it would be possible to achieve a global objective reliably without centralized coordination. In addition, the multi-agent system can be spatially distributed which is particularly useful in guarding, monitoring, and inspection applications. The purpose of this research is to implement a multi-agent system composed of state-of-art miniature mobile robots equipped with sensors and communication devices which support a wide range of applications including guarding, monitoring, and various, possibly hazardous, inspection tasks. In this research, a serious effort is made to bring mobile robots to real world applications in a cost-effective and a feasible way. In addition to coordinated cooperation, a physical multiagent system in which individual agents, or spatially distributed computational entities in general, interact laterally using simple communication medium, e.g. light, obeying inverse-power law, provides an opportunity to study the relationship of spatial patterns and the resulting dynamics. In this project we investigate how this relationship between spatial patterns and resulting dynamics could be utilized to create purposeful dynamic systems by spatially arranging computational entities in such a way that desired dynamics, in regard to some utility measure, emerges. The study of spatial patterns is motivated by the observation that nature has a remarkable ability to create complex patterns on various scales from physical matter. If these patterns are composed of entities that are able to interact with the environment, i.e. they are not closed systems; a dynamic system emerges in which interactions of the entities are inevitably regulated by the physical reality, including space. Assuming that space has an important role in defining the properties of a dynamic system composed of laterally interacting entities, it is possible to modify the dynamics of the system by altering the spatial pattern formed by the entities. It is clear that the patterns that exist in living organisms are to a large extent genetically determined. However, an infinite number of non-genetic spatial patterns exist on various scales in the physical world that form dynamic systems that are composed of laterally interacting entities -- both living and non-living. These patterns can be a result of physical forces acting in the physical reality, e.g. the gravity in start systems, or they can be a result of a complex stochastic selection process, i.e. adaptation, as in population dynamics and spatial ecology. The dramatic effect of space on dynamic systems composed of laterally interacting entities is perhaps best demonstrated by considering the tremendous effect of continental drift on the population dynamics in earth. The scientific goals of this project are: 1) To understand how a global objective can be achieved by a multiagent system without explicit regard to cooperation with the other agents. 2) To investigate the relationship of spatial patterns composed of laterally interacting entities (agents), and the resulting dynamics, and to study how this relationship could be utilized. 3) To investigate how humans can effortlessly interact and remote control a multi-agent system and to have meaningful information about the environment through it. 4) To study what are the minimal requirements for an agent and its sensor and communication capabilities to produce useful behaviors at the system level. This research is strongly connected to our earlier work on robot control systems, adaptation, and learning. However, the previous work was essentially based on simulations or a single real robot. The real physical system will provide us an opportunity to validate our results with a real multi-agent system and to bring the results to real world applications.

Contact information

juha . roning (at) ee . oulu . fi

[http://www.ee.oulu.fi/research/robotics/redi/]

[This page is CSS2 enabled. Your browser might not fully support it]