A Decentralized Mechanism for Distributed Coordination & Control

Speaker: Naftaly Minsky

16th October 2003 , 1.00 pm , Room 519 Claremont Tower

Abstract

It is my thesis that for a group of autonomous agents to interoperate effectively, they must be able to trust each other to comply with some common rules-of-engagement, or a policy. Moreover, it stands to reason that if the members of the group in question are heterogeneous, with little or no trust in each other, then their interaction-policy needs to be enforced; and that this enforcement needs to be de-centralized, if the group might be large.

Guided by this thesis we have developed a coordination mechanism called Law-Governed Interaction (LGI), that enables a community C of distributed agents to interact under an explicit and strictly enforced policy, called the ``law'' of this community.

This mechanism, which is currently prototyped by the Moses toolkit, has the following characteristics: (a) The membership of C can change dynamically, and can be very large. (b) LGI makes no assumptions about the structure and behavior of members of C, which can, therefore, be heterogeneous. (c) The deployment of a community under a specified law is easy, incremental, and can be done dynamically. (d) The enforcement of laws under LGI is strictly decentralized---for scalability. (e) LGI provides for flexible interoperability between policies.

In the talk, I will attempt to motivate this mechanism, describe its nature, and outline some of its applications.