interactive artificial learning in multi-agent systems

1
Interactive Artificial Learning in Multi-agent Systems Yomna M. Hassan, Salman Ahmed, and Jacob W. Crandall Computing and Information Science Program at the Masdar Institute of Science and Technology, Abu Dh Email: {yhassan, sahmed, jcrandall}@masdar.ac.ae On-going Research Multi-agent learning algorithms for coordination in smart power grids In power systems with renewable energy resources, demand response programs can be used to are encourage more efficient use of energy resources. Intelligent devices can be developed to help users respond effectively. One method we are considering for these devices is interactive evolutionary learning, wherein human input is provided to a genetic algorithm. We are developing interactive evolutionary algorithms that learn successfully in multi-agent systems with minimal human input. Basic structure of the algorithm is shown in the figure to the left. Learning By Demonstration in Repeated stochastic games We have performed preliminary investigating the usefulness of LbD in MAS. The simulation have been done on a repeated stochastic games based which models the iterative prisoner’s dilemma..Results show that LbD helps learning agents learn non-myopic equilibrium in repeated stochastic games when human demonstrations are well-informed. On the other hand, when human demonstrations are less informed, these agents sometimes learn behavior that produces (less-successful) myopic behavior. Task Scheduling in Multi-Vehicle Transportation Systems In general, machine learning algorithms rely heavily on the configuration stage, wherein the programmer selects relevant features and a distance metric. We are investigating the possibility of deriving the distance metric from interactions between the agent and the user. In particular, we are adapting and extending the CBA algorithm (Chernova and Veloso, 2009) for an online taxi problem. initialization Evaluate fittnes Check term ination conditiion criteria selection no crossover mutation New population terminate yes H um an input H um an input Introduction Many real-world problems, in which intelligent machines can be useful, require interactions between multiple intelligent agents. To overcome the challenges of previously used methods, we are adapting Interactive artificial learning (IAL) as a learning methodology in multi-agent systems (MAS). Learning by demonstration (LbD) and reward reinforcement have been studied previously in single agent environments. We are focusing on MAS. Interactive artificial learning process (C hernova and Veloso,2009) C onfigure N eeds Demonstration Act NO Train Yes N eeds H elp to U pdate Policy Update Policy NO U pdate Policy Yes M odification Planned ProgrammerInput Required U serInputR equired W ork Autonom sly Configure Plan A ct O bserve and Rew ard U pdate End U ser Step 1 Step 2 Step 3 Step 4 Step 5 Learning By D em onstration R ew ard R einforcem ent

Upload: yomna-ibrahim-hassan

Post on 02-Nov-2014

107 views

Category:

Technology


3 download

DESCRIPTION

Interactive Artificial Learning in Multi-agent Systems

TRANSCRIPT

Page 1: Interactive Artificial Learning in Multi-agent Systems

Interactive Artificial Learning in Multi-agent Systems Yomna M. Hassan, Salman Ahmed, and Jacob W. CrandallComputing and Information Science Program at the Masdar Institute of Science and Technology, Abu Dhabi, UAE.

Email: {yhassan, sahmed, jcrandall}@masdar.ac.ae

On-going Research

Multi-agent learning algorithms for coordination in smart power grids

In power systems with renewable energy resources, demand response programs can be used

to are encourage more efficient use of energy resources. Intelligent devices can be developed

to help users respond effectively. One method we are considering for these devices is

interactive evolutionary learning, wherein human input is provided to a genetic algorithm. We

are developing interactive evolutionary algorithms that learn successfully in multi-agent systems

with minimal human input. Basic structure of the algorithm is shown in the figure to the left.

Learning By Demonstration in Repeated stochastic games

We have performed preliminary investigating the usefulness of LbD in MAS. The simulation have been done on a repeated stochastic

games based which models the iterative prisoner’s dilemma..Results show that LbD helps learning agents learn non-myopic equilibrium in

repeated stochastic games when human demonstrations are

well-informed. On the other hand, when human demonstrations

are less informed, these agents sometimes learn behavior that

produces (less-successful) myopic behavior.

Task Scheduling in Multi-Vehicle Transportation Systems

In general, machine learning algorithms rely heavily on the configuration stage,

wherein the programmer selects relevant features and a distance metric. We are

investigating the possibility of deriving the distance metric from interactions between

the agent and the user. In particular, we are adapting and extending the CBA

algorithm (Chernova and Veloso, 2009) for an online taxi problem.

initialization Evaluate fittnesCheck termination conditiion criteria

selection

no

crossover

mutation

New population

terminateyes

Human input

Human input

Introduction

Many real-world problems, in which intelligent machines

can be useful, require interactions between multiple

intelligent agents. To overcome the challenges of

previously used methods, we are adapting Interactive

artificial learning (IAL) as a learning methodology in

multi-agent systems (MAS). Learning by demonstration

(LbD) and reward reinforcement have been studied

previously in single agent environments. We are

focusing on MAS.

Interactive artificial learning process

( Chernova and Veloso, 2009)Configure

Needs Demonstration

ActNO

Train

Yes

Needs Help to Update Policy

Update Policy

NO

Update Policy

Yes

Modification Planned

Programmer Input Required

User Input Required

Work Autonomsly

Configure Plan ActObserve

and Reward

Update

End User

Step 1 Step 2 Step 3 Step 4 Step 5

Learning By Demonstration

Reward Reinforcement