Download Objective Alignment in Multiagent Systems - Lecture Slides | ME 538 and more Study notes Mechanical Engineering in PDF only on Docsity! Kagan Tumer, kagan.tumer@oregonstate.edu ME 538 Fall 07: Agents and Multiagent Systems Week 3, Lecture 1: Collectives: Objective Alignment in Multiagent Systems Kagan Tumer, kagan.tumer@oregonstate.edu Collectives Relation to Other Fields Mechanism Design Reinforcement Learning Multi-Agent Systems Game Theory Swarm Intelligence Econophysics Computational Economics Operations Research • Collectives and the Design of Complex Systems, K. Tumer and D. Wolpert, editors. Springer, New York, 2004. ISBN 0-387-40165-2 Adaptive Control Kagan Tumer, kagan.tumer@oregonstate.edu • A collective is a large multi agent system where – Each agent has a private utility it is trying to optimize; and – A system utility function measures the full system’s performance • Important issues: – How to set private utility functions? – How to update them (team formation)? – Can agents compute those utility functions? – What happens when information is missing? – What happens when some agents start to fail? Collectives: Distributed Learning & Control Kagan Tumer, kagan.tumer@oregonstate.edu • System utility Valuation of company • Agents Employees • Private Utilities Compensation packages • Design problem (faced by the board): – How to set/modify compensation packages (private utilities) of the agents to increase valuation of company (system utility) • Salary/bonus • Benefits • Stock options – Note: Board does not tell each individual what to do. They set the “incentive packages” for employees (including the CEO). Analogy: A company Kagan Tumer, kagan.tumer@oregonstate.edu Learnability Low Learnability High Learnability High Learnability G(z) g i (z) Change in gi as a result of i’s actions Change in gi as a result of other agents’ actions "gi= Learnability: Degree to which an agent’s objective function is sensitive to its own actions, as opposed to the “background” noise of other agents’ actions Kagan Tumer, kagan.tumer@oregonstate.edu Properties G(z) High Factoredness Low Learnability Low Factoredness High Learnability High Factoredness High Learnability g i (z) Kagan Tumer, kagan.tumer@oregonstate.edu General Solution • To get utilities with high factoredness and learnability, start with: ! gi(z) =G(z) "G(z"i+ci) gi is aligned with G G(z-i+ci) is independent of i gi has cleaner signal than G G(z-i+ci) removes noise ! "gi(z) "z i = "G(z) "z i • If g, G differentiable, then: ! "G(z#i + ci) "z i = 0