Dynamic Agent Population in Agent-Based Distance Vector Routing Page: 5
6 p.View a full description of this paper.
Extracted Text
The following text was automatically extracted from the image on this page using optical character recognition software:
200 200
Initial Population = 20 Initial Population = 20
Initial Population= 20 ---- Initial Population= 20
180 InitialPopulation= 150 - 180 Initial Population= 150
Initial Population= 200 Initial Population= 200
160 160
140 140
120 120
100 - 100
80 80
60 60
40 40
20 20
0 0
50 100 150 200 0 50 100 150 200
Time (ms) Time (ms)
(a) Variance in Agent Population with I = 0.8, = 0.3, (b) Degree of Volatility A = 0.5
and A 1
Figure 2: Dynamic Control of Agent Population
kills itself. However if the Node Pheromone value reduces below a Cloning Threshold (1), the agent clones
itself. Before leaving n,, the agent deposits additional Node Pheromone at time t2. This approach controls
the agent population based on the inter-agent arrival time expressed as a function of the Node Pheromone.
If the inter-agent arrival time is small (e-x(At) > I) and the agent produced no updates in the existing
routing table entries, it implies an excessive number of agents in the system. On the other hand, if the
inter-agent arrival time is large (e-x(At) < 1), it implies there are a sub-optimal number of agents in the
system. However if 1 < e-x(At) < , the agent neither clones nor kills. Killing requires the agent to
destroy its instance along with its data segments. Cloning requires the agent to create another instance of
itself with same attributes and privileges.
Figure 2(a) shows the convergence of the agent population in a 40 node network with average degree
of 7. The results assume the I = 0.8, 1 = 0.3, and A = 1. It was observed that irrespective of the initial
population, the system converges to a fairly constant number of agents in the system (approximately 20).
Networks initialized with a small number of agents escalate the agent population to a certain value thereby
improving the network performance. However, the escalation of agent population occurs at every node
flooding the network with excessive agents. Nevertheless, the system realizes the overhead of large number
of agents and adjusts itself to dynamically reduce the population. On the other hand, networks initialized
with a large number of agents realize the per-agent overhead and continuously reduce the population until
it reaches a somewhat constant number.
Figure 2(b) displays the variance in agent population with Node Pheromones having reduced degree of
volatility (A). A reduced degree of volatility denote more stable pheromones. Hence their strength decay
at lower rates reducing the frequency of cloning functions and increasing the frequency of killing functions.
Low values of A significantly reduce the initial burst of agents in systems initialized with small agent
population. Although less volatile pheromones help to reduce the initial agent overhead, it also reduces
the sensitivity of the system. ADVR implementing a dynamic agent population may start with a single
agent or an arbitrary number of agents. Nevertheless, the agents coordinate themselves and converge to a
particular range of population. Although this range may not represent the ideal number of agents required
for the system to converge, it represents an optimal population based on the availability of resources. This
range however depends on the values of P, Q, and A. An adaptive system should adjust these values
dynamically based on its resource availability. Dynamically adjusting these thresholds in a distributed
fashion as a function of network topology, congestion level, etc. is a non-trivial problem and is beyond the
scope of this paper.
Upcoming Pages
Here’s what’s next.
Search Inside
This paper can be searched. Note: Results may vary based on the legibility of text within the document.
Tools / Downloads
Get a copy of this page or view the extracted text.
Citing and Sharing
Basic information for referencing this web page. We also provide extended guidance on usage rights, references, copying or embedding.
Reference the current page of this Paper.
Amin, Kaizar A. & Mikler, Armin R. Dynamic Agent Population in Agent-Based Distance Vector Routing, paper, August 2002; (https://digital.library.unt.edu/ark:/67531/metadc132968/m1/5/: accessed April 25, 2024), University of North Texas Libraries, UNT Digital Library, https://digital.library.unt.edu; crediting UNT College of Engineering.