An important and di cult problem in the management of telecommunications networks is how good policies can be developed for admitting or blocking customer call re quests for a network of limited server capacity This paper describes a new co evolutionary approach to this problem which describes customers and servers as indi viduals within a technological ecology and models their interaction using a modi ed version of the Iterated Pris oner s Dilemma IPD The model is also deployed to tackle the more complicated class problem in which there are two customer populations each with di erent priorities and service requirements For both classes of problem good results are achieved in terms of low call blocking rates even in the face of system noise and server failure Although the present research is exploratory in char acter it is proposed that this co evolutionary approach could eventually be deployed in real time in a telecom munications context allowing a network to adapt auto matically to changes and problems as they arise More immediately it is suggested that this kind of system could be used in simulation o line and combined with some degree of executive control in a hybrid system Such a technology would be within reach of current en gineering viability
[1]
Thomas L. Magnanti,et al.
Network Design and Transportation Planning: Models and Algorithms
,
1984,
Transp. Sci..
[2]
R. May,et al.
Stability and Complexity in Model Ecosystems
,
1976,
IEEE Transactions on Systems, Man, and Cybernetics.
[3]
Mark D. Smucker,et al.
Iterated Prisoner's Dilemma with Choice and Refusal of Partners: Evolutionary Results
,
1995,
ECAL.
[4]
Kristian Lindgren,et al.
Evolutionary phenomena in simple dynamics
,
1992
.
[5]
L. Tesfatsion,et al.
Preferential partner selection in an evolutionary study of Prisoner's Dilemma.
,
1994,
Bio Systems.
[6]
W. Hamilton,et al.
The evolution of cooperation.
,
1984,
Science.
[7]
Michael L. Littman,et al.
A Distributed Reinforcement Learning Scheme for Network Routing
,
1993
.