Simulation for determining initial system effect for coordination in multiagent systems Murat Aydin

1
Simulation for determining initial system effect
for coordinationin multiagent systems Murat
Aydin
2
What is coordination

• Coordination is a central issue in multiagent
  systems. It is globally coherent behavior of the
  agents.,
• Coordination is required to achieve a goal in
  multiagent systems

3
Problem

• Initial system status means number of agents,
  number of resources, distribution of agent on
  resources and distribution of resources on
  locations.
• initial system state must have some effects on
  coordination effort which is not explored very
  well in literature

4
Problem cont.

• Resource change effect is the same with initial
  system status effect. Because adding resources
  makes the system uncoordinated and some
  coordination is needed. System prior to
  coordination is initial system.

5
Motivation

• Identifying effects of initial system on
  coordination allows system designer to change
  resources in a manner in which coordination can
  take less time. Also knowing effect of initial
  system status allows feasibility researches i.e.
  the system will complete the or the system will
  always be in a challenge with coordination.

6
Aim

• how coordination is affected by an increase in
  number of resources per agent.
• how resource change patterns affect coordination

7
Examples

• Example 1 Some physical jobs i.e. Pollen
  collection by bee agents
• Example 2 Nonhomogeneous resources
• Example 3 Service selection where quality of
  service is dependent on number of clients and
  coordination is getting best service.

8
DefinitionsResource Change Distribution

Figure 1.a Before change Figure 1.b After
change
9
Before and after coordination
Number of Agents per resource

Figure 2.a before change Figure 2.b after
change, after coordination
10
DefinitionsRelative Resource Change Distribution

• resource changes are relative to agent
  distribution over resources.

After, a second resource change (3,9,15,21,12) is
applied. The resource change viewed by agents

Figure 3.a second resource change Figure 3.b
resource change viewed by agents
11
Experimental setup
Experimental setup
Knowledge (Kn), Choice (Ch) and knowledge sharing
of agents at resources i and j Singh and
Rustogi
12
DefinitionsKey factors of coordination

• Knowledge
• Choice
• Shared Knowledge
• Inertia
• Tolerance to imprecision

13
DefinitionsDecision Protocol
Decision Protocol from Singh and Rustogi
14
SimulationPerformance Metric

• Distance from coordination Number of movements
  required to coordinate

15
Simulation

• control and test systems with the same initial
  status
• For test system experimentally or normally
  distributed resource change will be applied.
• For the control system, same number of resource
  change will be applied with a uniform
  distribution.

16
Evaluation

• apply statistical analysis to test whether the
  difference in mean number of steps to coord for
  each group is significant or not.
• If different,analyze results to come up with an
  explanation about this difference
• try to conclude with a formula which lets to
  predict number of movements

17
References

• Singh and Rustogi Sudhir K. Rustogi and
  Munindar P. Singh. Be Patient and Tolerate
  Imprecision How Autonomous Agents can Coordinate
  Effectively.
• Candale and Sen Teddy Candale and Sandip Sen.
  Fast convergence to satisfying distributions.

18
THANKS