TY - GEN
T1 - Towards distributing agent intelligence
T2 - 13th International Conference on Principles and Practice of Multi-Agent Systems, PRIMA 2010
AU - Balsdon, Quintin J.
AU - Ehlers, Elize M.
PY - 2012
Y1 - 2012
N2 - Autonomous agents are often restricted by the programs that make up their 'intelligence' because they are installed on the same hardware as the agent. Since intelligence is software and therefore abstract, it is possible to separate the components which 'create' the agent's intelligence from the agent itself. The disembodiment of intelligence allows agents to access components that may not be suited to their hardware for physical reasons, such as storage capacity or computational complexity. It has been long established that humans find solutions to problems by dividing a problem into a series or smaller sub-problems [1; 2]. Using web services, 'intelligence components' can be created which perform a simple generic task on behalf of a client agent. These components may be used in different combinations in order to create customized solutions for particular problems. Intelligence components may be distributed across servers in different locations, allowing other agents to benefit from the differing implementations. In addition, software may be updated remotely by updating individual components. The model is aimed at creating a repository of useful functionality which may enable intelligent agents to focus on the process of intelligence rather than processing individual environmental states. The solution presented demonstrates that an agent may access distributed components in order to control behaviour, taking into consideration that components themselves have no concept of the environment in which the requesting agent exists. The problem of translating a 'model' solution into the environment-specific solution is therefore left to the agent.
AB - Autonomous agents are often restricted by the programs that make up their 'intelligence' because they are installed on the same hardware as the agent. Since intelligence is software and therefore abstract, it is possible to separate the components which 'create' the agent's intelligence from the agent itself. The disembodiment of intelligence allows agents to access components that may not be suited to their hardware for physical reasons, such as storage capacity or computational complexity. It has been long established that humans find solutions to problems by dividing a problem into a series or smaller sub-problems [1; 2]. Using web services, 'intelligence components' can be created which perform a simple generic task on behalf of a client agent. These components may be used in different combinations in order to create customized solutions for particular problems. Intelligence components may be distributed across servers in different locations, allowing other agents to benefit from the differing implementations. In addition, software may be updated remotely by updating individual components. The model is aimed at creating a repository of useful functionality which may enable intelligent agents to focus on the process of intelligence rather than processing individual environmental states. The solution presented demonstrates that an agent may access distributed components in order to control behaviour, taking into consideration that components themselves have no concept of the environment in which the requesting agent exists. The problem of translating a 'model' solution into the environment-specific solution is therefore left to the agent.
KW - Coordination and Concurrency Runtime
KW - Distributed Software Services
KW - Microsoft Robotics Developer Studio
KW - Serialization
UR - http://www.scopus.com/inward/record.url?scp=84874652085&partnerID=8YFLogxK
U2 - 10.1007/978-3-642-25920-3_24
DO - 10.1007/978-3-642-25920-3_24
M3 - Conference contribution
AN - SCOPUS:84874652085
SN - 9783642259197
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 340
EP - 354
BT - Principles and Practice of Multi-Agent Systems - 13th International Conference, PRIMA 2010, Revised Selected Papers
Y2 - 12 November 2010 through 15 November 2010
ER -