Kasbah: An Agent MarketPlace for Buying and Selling goods

1
Kasbah An Agent MarketPlace for Buying and
Selling goods
Mohan DeSouza Presentation III -
10/30/2002 Control of Autonomous
Agents University of Memphis, TN
2
Work of ..

• Anthony Chavez (1996)
• MIT Media Lab, Cambridge, MA
• Patty Maes (1996)
• MIT Media Lab, Cambridge, MA

3
Introduction

• Automate Buying/Selling/Negotiation on the
  Internet
• Reactive Agent

4
Problem Description

• Seller who want to sell
• Buyers who want to buy
• Haggling/Negotiation
• Buyers/Sellers get poor dealslack of time,search
  limitation,lack of information

5
The SolutionKasbah

• KasbahTraditional fortress and palace of towns
  in pre-colonial North Africa.
• Kasbah-online marketplace
• Users create buying and selling agents which
  negotiate on their behalf
• Multi-agent system

6
Selling Agents

• User Provide Control
• Desired date to sell the item by
• Deisred price
• Lowest Acceptable price
• Negotiation Strategy (decay functionlinear/quadra
  tic/cubic)
• Get user approval before finalizing deal
• Send email notification when agreement reached.

7
Buyer Agent

• User Provide Control
• Desired date to buy the item by
• Desired price
• Highest Acceptable price
• Negotiation Strategy (decay functionlinear/quadra
  tic/cubic)
• Get user approval before finalizing deal
• Send email notification when agreement reached.

8
The Marketplace

• Place where Buying and Selling agents meet.
• tents of interest are formed according to the
  product to be bought/sold
• Marketplace informs buyers/sellers of each other

9
Control of Agents

• Methods Implemented by both agents
• accept-offer?(agent,from-agent,offer)
• what-is-price?(agent,from-agent)
• what-is-item?(agent,from-agent)

10
Marketplace-Agent Interface

• Add-potential-customers(sell-agent,potential-custo
  mers)
• Add-potential-sellers(buy-agent,
  potential-sellers)
• Agent-terminated(marketplace,agent)
• Deal-made(marketplace,sell-agent,buy-agent,item,pr
  ice)
• do-thing method

11
Control of Agents contd

• Marketplace runs in cycles
• Each agent is given a chance to do_thing per
  cycle
• Agent
• Determine the current asking price
• Decide which agent to talk to
• Talk to the potential contact.

12
Determine the current asking price

• 1st dayasking pricedesired price
• Any day in betweenasking price is raised/lowered
  according to the decayed function.
• Last day asking pricelowest price

13
Decide which agent to talk to

• Talk to those agents which seem the most
  promising
• Those who have never been spoken to
• If all agents have been contacted-then pick the
  one whose last known offering price is the
  highest/lowest (those agents which have indicated
  a willingness to pay a higher (or sell for a
  lower) price.

14
Talk to the potential contact

• Agent offers to buy/sell item at asking price. If
  the contacted agent accepts the job is done.
• If the contacted agent rejects the offer, then it
  is asked what its offering price is (this price
  is recorded)

15
The Agents Environment

• The Internet

16
The Agents Sensory Capability

• Senses text

17
Agents Primitive Actions

• accept-offer?(agent,from-agent,offer)
• what-is-price?(agent,from-agent)
• what-is-item?(agent,from-agent)
• Add-potential-customers(sell-agent,potential-custo
  mers) or Add-potential-sellers(buy-agent,
  potential-sellers)
• Deal-made(marketplace,sell-agent,buy-agent,item,pr
  ice)

18
Agents Action Selection Mechanism

• Already explained

19
References

• Chavez, A.,Maes, P. Kasbah An Agent Marketplace
  for Buying and Selling Goods
• Chavez, A.,Dreilinger, D., Guttman R., Maes, P.
  A Real-Life Experiment in Creating an Agent
  Marketplace