Auctions

1
Auctions
Automated Negotiation
2
Auction Protocols

• English auctions
• First price sealed-bid auctions
• Second best price sealed-bid auctions (Vickery
  auctions)
• Dutch auctions

3
(No Transcript)
4
The Contract Net

• R. G. Smith and R. Davis

5
DPS System Characteristicsand Consequences

• Communication is slower than computation loose
  coupling efficient protocol modular
  problems problems with large grain size

6
More DPS System Characteristicsand Consequences

• Any unique node is a potential bottleneck
  distribute data distribute control organized
  behavior is hard to guarantee
  (since no one node has complete picture)

7
The Contract Net

• An approach to distributed problem solving,
  focusing on task distribution
• Task distribution viewed as a kind of contract
  negotiation
• Protocol specifies content of communication,
  not just form
• Two-way transfer of information is natural
  extension of transfer of control mechanisms

8
Four Phases to Solution,as Seen in Contract Net

• 1. Problem Decomposition
• 2. Sub-problem distribution
• 3. Sub-problem solution
• 4. Answer synthesis

The contract net protocol deals with phase 2.
9
Contract Net

• The collection of nodes is the contract net
• Each node on the network can, at different times
  or for different tasks, be a manager or a
  contractor
• When a node gets a composite task (or for any
  reason cant solve its present task), it breaks
  it into subtasks (if possible) and announces them
  (acting as a manager), receives bids from
  potential contractors, then awards the job
  (example domain network resource management,
  printers, )

10
Node Issues Task Announcement
Task Announcement
Manager
11
Idle Node Listening toTask Announcements
Manager
Potential Contractor
Manager
Manager
12
Node Submitting a Bid
Bid
Manager
Potential Contractor
13
Manager listening to bids
Bids
Potential Contractor
Manager
Potential Contractor
14
Manager Making an Award
Award
Manager
Contractor
15
Contract Established
Contract
Manager
Contractor
16
Domain-Specific Evaluation

• Task announcement message prompts potential
  contractors to use domain specific task
  evaluation procedures there is deliberation
  going on, not just selection perhaps no tasks
  are suitable at present
• Manager considers submitted bids using domain
  specific bid evaluation procedure

17
Types of Messages

• Task announcement
• Bid
• Award
• Interim report (on progress)
• Final report (including result description)
• Termination message (if manager wants to
  terminate contract)

18
Efficiency Modifications

• Focused addressing when general broadcast
  isnt required
• Directed contracts when manager already knows
  which node is appropriate
• Request-response mechanism for simple
  transfer of information without overhead of
  contracting
• Node-available message reverses initiative of
  negotiation process

19
Message Format

• Task Announcement Slots Eligibility
  specification Task abstraction Bid
  specification Expiration time

20
Task Announcement Example(common internode
language)

• To
• From 25
• Type Task Announcement
• Contract 436
• Eligibility Specification Must-Have FFTBOX
• Task Abstraction
• Task Type Fourier Transform
• Number-Points 1024
• Node Name 25
• Position LAT 64N LONG 10W
• Bid Specification Completion-Time
• Expiration Time 29 1645Z NOV 1980

21

• The existence of a common internode language
  allows new nodes to be added to the system
  modularly, without the need for explicit linking
  to others in the network (e.g., as needed in
  standard procedure calling).

22
Applications of the contract Net

• Sensing
• Task Allocation (Malone)
• Delivery companies (Sandholm)
• Market-oriented programming (Wellman)

23
Bidding Mechanisms for Data Allocation

• A user sends its query directly to the server
  where the needed document is stored.

24
Environment Description
a client
a document
a query
distance
serverj
serveri
area i
area j
25
Utility Function

• Each server is concerned only whether a dataset
  is stored locally or remotely, but is indifferent
  with respect to different remote location of the
  dataset.

26
The Trading Mechanism

• Bidding sessions are carried on during predefined
  time periods.
• In each bidding session, the location of the new
  datasets is determined and the location of each
  old dataset can be changed.
• Until a decision is reached, the new datasets are
  stored in a temporary buffer.

27
The Trading Mechanism - cont.

• Each dataset has an initial owner (called
  contractor(ds)), according to the static
  allocation.
• For an old dataset - the server which stores it.
• For a new dataset - the server with the nearest
  topics (defined according to the topics of the
  datasets stored by this server).

28
The Bidding Steps

• Each server broadcasts an announcement for each
  new dataset it owns, and also for some of its old
  local datasets.
• For each such announcement, each server sends the
  price it is willing to pay in order to store the
  dataset locally.
• The winner of each dataset is determined by its
  contractor. It broadcasts a message, including
  the winner, the price it has to pay, and the
  server which bids this price.

29
Cost of Reallocating Old Datasets

• move_cost(ds,bidder)the cost for contractor(ds)
  for moving ds from its current location to
  bidder. (for new datasets, move_cost0)
• obtain_cost(ds,bidder)the cost for bidder for
  moving ds from its current location to bidder.

30
Protocol Details

• winner(ds) denotes the winner of dataset ds.
• winner(ds) arg max bidder move(ds)true
  price_suggested(bidder,ds) -
  move_cost(ds,bidder) none otherwise

31
Protocol Details - cont.

• price(ds) denotes the price paid by the winner
  for dataset ds.
• price(ds) second_max bidder?SERVERS
  price_suggested(s,ds) -move_cost(ds,bidder)
  move_cost(ds,winner).

32
Bidding Strategies

• Attributemove(ds)true ifsecond_max
  bidder?SERVERS
• price_suggested(bidder,ds) -
• move_cost(ds,bidder) ³ Ucontractor(ds)(ds,con
  tractor(ds)).

33
Bidding Strategies - cont.

• LemmaIf the winner server had bid its true value
  of storing the dataset locally, then it will
  have a nonnegative utility from obtaining it.
• LemmaEach server will bid its utility from
  obtaining the datasetprice_suggested(bidder,ds)
  Ubidder(ds,bidder) -
  obtain_cost(ds,bidder).

34
Bidding Strategies - cont.

• TheoremIf announcing and bidding are free, then
  the allocation reached by the bidding protocol
  leads to better or equal utility for each server
  than does the static policy.The utility
  function is evaluated according to the expected
  profits of the server from the allocation.

35
Usage Estimation

• Each server knows only the usage of datasets
  stored locally.
• For new datasets and remote datasets, the server
  has no information about past usage.
• It estimates the future usage of new and remote
  datasets, using the past usage of local datasets,
  which contain similar topics.

36
Queries Structure

• We assume that a query sent to a server contains
  a list of required documents.
• This is the situation if the search mechanism to
  find the required documents is installed locally
  by the client.
• In this situation, the server has to learn from
  the queries about its local documents, to the
  expected usage of other documents, in order to
  decide whether it needs them or not.

37
Usage Prediction

• We assume that a dataset contains several
  keywords (k1..kn).
• For each local dataset ds, and each server d, the
  server saves the past usage of ds by d, in the
  last period
• Then, it has to predict the future usage of ds by
  d. It assumes the same behavior than in the past.

38
Usage Prediction - cont.

• It is assumed that the users are interested in
  keywords, so the usage of a dataset is a function
  of the keywords it contains.
• The simplest model is when a dataset usage is
  the sum of the the usage of each of its
  keywords. However, the relationship between the
  keywords and the dataset may be different.

39
Usage Prediction - cont.

• The server has to learn about usage of datasets
  not stored locally
• We suggest that it will build a Neural Network
  for learning the usage template of each area.

40
What is Neural Network

• A neural network is composed of a number of
  nodes, or units, connected by links.
• Each link has numeric weight associated with it.
• The weight are modified so as to try to bring the
  networks input/output behavior more into line
  with that of the environment providing the input.

41
Neural Network - Cont.
Output unit
Output layer
Hidden layer
Input layer
Input unit
42
Structure of the Neural Network

• For each area d, we build a neural network.
• Each dataset stored by the server in area d, is
  one example for the neural network of d.
• The inputs of the examples contain, for each
  possible keyword, whether it exist in this
  dataset, or not.

43
Structure of the Neural Network - cont.

• The output unit of the Neural Network for area d,
  is its past usage of this dataset.
• In order to find the expected usage of another
  dataset, ds2, by d, we provide the network with
  the keywords of ds2.
• The output of the network is its predicted usage
  of ds2 by area d.

44
Structure of the NN
Output unit the usage of the dataset by a
certain area.
Hidden layer
For a certain dataset, for each keyword k there
is an input unit 1 if the dataset contains k. 0
otherwise.
45
Experimental Evaluation - Results Measurement

• vcosts(alloc) - the variable cost of an
  allocation, which consists of the transmission
  costs due to the flow of queries.
• vcost_ratio the ratio of the variable costs when
  using the bidding mechanism and the variable
  costs of the static allocation.

46
Experimental Evaluation

• Complete information concerning previous queries
  (still uncertainty)
• The bidding mechanism reaches results near to
  that of the optimal allocation (reached by a
  central decision maker).
• The bidding mechanism yields a lower standard
  deviation of the servers utilities than the
  optimal allocation.
• Incomplete information
• The results of the bidding mechanism are better
  than those of static allocation.

47
Influence of ParametersComplete Information, no
movements of old datasets

• As the standard deviation of the distances
  increases, vcost_ratio decreases.

48
Influence of Parameters - cont.

• When increasing the number of servers and the
  number datasets, vcost_ratio is not influenced.
• query_price, answer_cost, storage_cost,
  dataset_size and retrieve_ cost do not influence
  vcost_ratio.
• usage, std. usage, distance do not influence
  vcost_ratio.

49
Influence of Learning on the System

• As epsilon decreases, vcost ratio increases the
  system behaves better.

50
Conclusion

• We have considered the data allocation problem in
  a distributed environment.
• We have presented the utility function of the
  servers, which expresses their preferences over
  the data allocation.
• We have proposed using a bidding protocol for
  solving the problem.

51
Conclusion - cont.

• We have considered complete as well as incomplete
  information.
• For the complete information case, we have proved
  that the results obtained by the bidding
  mechanism are better than those of the static
  allocation, and closed to the optimal results.

52
Conclusion - cont.

• For the incomplete information environment We
  have developed a neural-network based learning
  mechanism.
• For each area d, we build a neural network,
  trained by the server of d.
• By this network, we find expectation for other
  datasets, not currently stored by d.
• We found, by simulation, that the results
  obtained are still significantly better than the
  static allocation.

53
Future Work

• Future Work
• Datasets can be stored in more than one server.
• Bounded rationality.
• Repeated game.

54
Reaching Agreements Through Argumentation

• Collaborator Katia Sycara, Madhura Nirkhe, Amir
  Evenchik, and
• Ariel Stolman

55
Introduction

• Argumentation--an iterative process emerging from
  exchanges among agents to persuade each other and
  bring about a change in intentions.
• A logical model of the mental states of the
  agents beliefs, desires, intentions, goals.
• The logic is used to specify argument formulation
  and a basis for Automated Negotiation Agent.

56
Agents as Belief, Desire, Intention systems

• Belief
• information about the current world state
• subjective
• Desire
• preferences over future world states
• can be inconsistent (in contrast to goals)
• Intentions
• set of goals the agent is committed to achieve
• the agents runtime stack
• Formal models mostly modal logics with
  possible-worlds semantics

57
Logic Background

• Modal logics Kripke structures
• Syntactic Approaches
• Baysen Networks

58
Modal Logics
A

• Language there is a set of n agents Kia
  --- i knows a Bia ----i believes a P -- a
  set of primitive propositions P,QExamples K1a
  
• Semantics A Kripke Structure consists of four
  elements
• A set of possible worlds
• p (w) P----- True,False
• n binary relations on the worlds 1, 2,.

59
Example of a Kripke Structure

• Ww1,w2,w3 M,w1 p K1p

w1
w2
P,Q
Q
1
2
2
1
P
w3
60
Axioms

• Kia Ki(a--b) --- Kib
• If -a then -Kia
• Kia --a
• Kia- KiKia
• Kia -- Ki Kia
• Bi false
• Each axiom can be associated with a condition on
  the binary relation.

61
Problems in using Possible Worlds Semantics

• Logical omniscience--the agent believes all the
  logical consequences of its belief.
• The agent believes in all tautologies.
• Philosophers possible worlds do not exist.

62
Minimal Models partial solution

• The intension of a sentence the set of possible
  worlds in which the sentence is satisfied
• Note if two sentences have the same intensions
  then they are semantically equivalent.
• A sentence is a belief at a given world if its
  intension is belief-accessible.
• According to this definition, the agent's beliefs
  are not closed under inferences the agent may
  even believe in contradictions.

63
Minimal model example
P Q
P Q
P Q
P Q
64
Beliefs, Desires, Goals and Intentions

• We use time lines rather than possible worlds.
• An agent's belief set includes beliefs concerning
  the world and beliefs concerning mental states of
  other agents.
• An agent may be mistaken in both kinds of beliefs
  and beliefs may be inconsistent.
• The beliefs are used to generate arguments in the
  negotiations.
• Desires may be inconsistent.
• Goals a consistent subset of the set of desires.
  
• Intentions serves to contribute to one or more of
  the agent's desires.

65
Intentions

• Two types Intention-To and Intention-That
• Intention-to refer to actions that are within
  the direct control of the agent.
• Intention-that refer to propositions that are
  not directly within the agent's realm of
  control, that it must rely on other agents for
  satisfying-- can be achieved through
  argumentation.

66
Argumentation Types

• A promise of a future reward.
• A threat.
• An appeal to past promise.
• Appeal to precedents as counter example.
• Appeal to prevailing practice.
• Appeal to self-interests

67
Example 2 Robots

• Two mobile robots on Mars each built to maximize
  its own utility.
• R1 requests R2 to dig for a mineral. R2 refuses.
  R1 responds with a threat If you do not dig
  for me, I will break your antenna''. R2 needs to
  evaluate this threat.
• Another possibility R1 promises a reward If
  you dig for me today, I will help you move your
  equipment tomorrow.'' R2 needs to evaluate the
  promise of future reward.

68
Usage of the logic

• Specification for agent design the model
  constraints certain planning and negotiation
  processes. Axioms for argumentation types
• The logic is used by the agents themselves ANA
  (Automated Negotiation Agent)

69
ANA

• Complies with the definition of an Agent Oriented
  Programming (AOP) system (Shoham)
• The agent is represented using notions of mental
  states
• The agent's actions depend on these mental
  states
• The agent's mental state may change over time
• Mental state changes are driven by inference
  rules.

70
The Block World Environment
2
11
1 2 3 4 5?
71
Mental State Model

• Beliefs
• b(agent1,world_state(blockE / 5 / 1,blockD / 4 /
  1,blockC / 3 / 1,blockB / 2 / 1,blockA / 1 /
  1),0,2,t).
• Desires
• desire(desire1,0,agent1 ,blockB / 6 / 2, 39,0).
• desire(desire2,0,agent1, blockB / 1 / 1, 29,0).
• desire(desire3,0,agent1, blockB / 6 / 1, 35,0).
• desire(desire4,0,agent1, blockE / 2 / 1, 38,0).
• Goals
• goal(agent1, 0, blockB / 6 / 2 / desire1,
  blockE / 2 / 1 / desire4)

72
Mental State Model

• Desired World
• desired_world( blockC / 3 / 1 /unused_block,
  blockA / 1 / 1 /unused_block, blockD / 6 / 1
  /supporting, blockB / 6 / 2 /desire1,
  blockE / 2 / 1 /desire4).
• Intentions
• intention(1,agent1,0,that,intention_is_done(agent1
  ,0), blockB / 2 / 1 / 7 / 1,0,
  towards_goals).
• intention(2,agent1,0,to,intention_is_done(agent1,1
  ), blockD / 4 / 1 / 6 / 1,0, supporting).
• intention(3,agent1,0, that,intention_is_done(
  agent1,2), blockB / 7 / 1 / 6 / 2,0,
  desire1).
• intention(4,agent1,0,to,intention_is_done(agent1,3
  ), blockE / 5 / 1 / 2 / 1,0, desire4).

73
Agent Infrastructure Agent Life Cycle
74
The Agent Life Cycle Reading Messages

• Types of messages
• Queue
• Waiting for answers
• Negotiation and world change aspects
• Inconsistency recovery

75
The Agent Life Cycle Dealing with the agents
own threats

• Detection
• Make abstract threats concrete
• Execute evaluation

76
The Agent Life Cycle Planning next step
Planning next step

• Mental states usage
• Backtracking
• Better than current state
• New state or dead end
• Achievable plan

77
The Agent Life Cycle Performing next intention

• Intention to - intention that
• Other agent listening?
• One argument per cycle.

78
Agent Definition Examples

• Agent Type agent_type(robot_name,
  memory-less).
• Agent Capability capable(robot, blockC / 3 /
  1 / 4 / 1).
• Agent Beliefs b(first_robot, capable(second_robot
  , AnyAction),
  0, t).
• Agent Desiresdesire(first_desire, 0, robot,
  blockA/3/1, 15, 1).

79
Agent Infrastructure Agent Parameters List

• Cooperativeness
• Reliability (promises keeping)
• Assertiveness
• Performance threshold (Asynchronous action)
• Usage of first argument
• Argument direction
• Knowledge about other desires
• Knowledge about other capabilities
• Measurement of other agent promises keeping
• Execution of threats by another agent

80
A promise of a future reward

• Application conditions
• Opponent agent can perform the requested action.
• The reward action will help the opponent achieve
  a goal (requires knowledge of opponent desires).
• Argument not used in the near past.
• Implementation
• Generate opponents expected intentions.
• Offer one of the intentions as a reward
• Mutual intention which opponent cannot perform by
  itself (requires knowledge of opponent
  capabilities).
• Opponents intention which it cannot perform.
• Any mutual intention.
• Any Opponents intention.

81
A threat

• Application conditions
• Opponent agent can perform the requested action.
• The threat action will interfere with the
  opponents achieving some goals (requires
  knowledge of opponent desires).
• Argument not used in the near past.
• Implementation
• Agent chooses best cube (requires knowledge of
  opponent capabilities).
• Agent chooses best desire.
• Agent chooses a threshing action
• Moving out.
• Blocking.
• Interfering.

82
Request Evaluation Mechanism- Parameters List

• DL (Doing Length)
• NDL (Not Doing Length)
• DTL (Doing That Length)
• NDTL (Not Doing That Length)
• PL (Punish Length)
• PTL (Punish That Length)
• DP (Doing Preference)
• NDP (Not Doing Preference)

83
Request Evaluation Mechanism- Agent Parameters

• CP The agents cooperativness.
• AS The agents assertiveness.
• RL The agents reliability.
• ORL The Other agents reliability for keeping
  promises.
• OTE The Other agents percentage of threat
  executing.

84
Request Evaluation Mechanism- The Formulas
85
Request Evaluation Mechanism- The Formulas
86
Experiments Results

• Negotiating is better than not negotiating only
  where each agent has particular expertise.
• Negotiating is better than not negotiating only
  where the agents have complete information.
• Negotiating is better than not negotiating only
  for mutually cooperative agents or for an
  aggressive agent with a cooperative opponent.
• Environment (game time, resources) effects the
  negotiations results.

87
Negotiations vs. no negotiations

• When the agents that do not negotiate succeed in
  obtaining only 29.8 of their desires preference
  values, the negotiating agents succeed in
  obtaining 40.4, on the average. (F5.047,
  p

88
Complete information vs. no information

• Agents that had no information succeed in
  obtaining a success rate of only 30.8, while
  agents that had full information succeed in
  obtaining 40.4 on the average.
  (F4.326,p

89
Using the first argument vs. using the best found

• Agents that used the first argument succeed in
  obtaining a success rate of only 34.8, while
  agents that used the best argument succeed in
  obtaining 40.4 on the average, but this result
  is not significant. (F2.28,p

90
Cooperative vs. Aggressive agent
23.5 41.4 (F10.78,p
91
Cooperative and Aggressive Agents vs. No
Negotiations
38.3 29.8 (F6.01, p 92
No negotiations VS. Aggressive Negotiation
38.3 20.8 (F10.03, p 93
Cooperative vs. aggressive

• Aggressive vs. cooperative 41.4
• Two cooperatives 38.3
• No negotiation 29.8
• Cooperative vs. aggressive 23.5
• Two aggressive 20.8

94
Environment Constraints
Number of desires
95
Environment Constraints
Time for game
16.7 21.7 (F2.41, p 96
Is it worth it to use formal methods for
Multi-Agent Systems in general and Negotiations
in particular?
97
Game-theory Based Frameworks(Non-cooperative
Models)

• Strategic-negotiation model based on
  alternating offers model of rubinstein.
  Applications
• Data allocation (schwartz kraus AAAI97),
• Resource allocation , task distribution
  (kraus wilkenfeld zlotkin AIJ95,
  kraus AMAI97), hostage crisis (kraus wilkenfeld
  TSMC93).

98
Advantages and DifficultiesNegotiation on Data
Allocation

• Beneficial results proved to be better than
  current methods simple strategies.
• Problems
• Need to develop utility functions
• Finding possible action identifying optimal
  allocations is NP complete
• Incomplete information game-theory
  provides limited solutions.

99
Game-theory Based Frameworks(Non-cooperative
Models)

• Auctions applications
• Data allocation (schwartz kraus ATAL97),
• Electronic commerce.
• Subcontracting based on principle agent
  models. Applications
• Task allocation (kraus, AIJ96).

100
Advantages and DifficultiesAuctions for Data
Allocation

• Beneficial results proved to be better than
  current methods.
• Problems
• Utility functions,
• Applicable only when a server is concerned only
  about the data stored locally,
• Difficult to find bidding when there is
  incomplete information and the evaluations are
  dependant on each other no procedures.

101
Game-theory Based Frameworks(Cooperative Models)

• Coalition theories applications
• Group and teams formation (shehory kraus CI99).
• Benefits well-defined concepts of stability
  mechanisms to divide benefits.
• Difficulties utility functions, no procedures
  for coalition formation exponential problems.
• DPS model combinatory theories operations
  research (shehory kraus AIJ98).

102
Decision-theory Based Frameworks

• Multi-attributed decision making application
• Intentions reconciliation in SharedPlans
  (grosz kraus, 98).
• Benefits using results of MADM, e.G., Specific
  method is not so important, standardization
  techniques.
• Problems choosing attributes assigning values,
  choosing weights.

103
Logical Models

• Modal logic BDI modelsapplications
• Automated argumentation's (kraus, sycara
  eventchick AIJ99).
• Specification of sharedplans (grosz kraus
  AIJ96).
• Bounded agents (nirkhe, kraus, perlis JLC97).
• Agents reasoning about other agents (kraus
  lehmann TCT88 kraus subrahmanian IJIS95).

104
Advantages and DifficultiesLogical Models

• Formal models with well studied
  propertiesexcellent for specification.
• Problems
• Some assumptions are not valid (e.g.,
  omnicience).
• Complexity problems.
• There are no procedures for actions required a
  lot of programming decision making developing
  preferences.

105
Physics Based Models

• Physical models of particle-dynamics
  Applications Cooperation in large-scale
  multi-agent systems freight deliveries within a
  metropolitan area. (Shehory
  Kraus ECAI96 Shehory, Kraus Yadgar ATAL98).
• Benefits efficient inherits the physics
  properties.
• Problems adjustments potential functions

106
Summary

• Benefits formal models which have already been
  studied lead to efficient results. No need to
  invent the wheel.
• Problems
• Restrictions and assumptions made by game-theory
  are not valid in real world MAS situations
  extensions are needed.
• It is difficult to develop utility functions.
• Complexity problems.