This project contains extensions to NS-2
(versions 2.27 and 2.28) which enables the ACO behavior in WSNs. The
module is adaptable to the current version of NS-2 since it is installed as a
new routing protocol. Written mainly in C++ the ACO module (AntSense) produces
the main attributes, however it was design in order to be used by any other technology
supported by this simulator.
When designing a routing protocol for WSNs it is important to consider the
packet length of each type of packets used by our protocol, since, as already
explained communication is an energy-expensive function. In AntSense
development, several approaches were used concerning types of packets and
different lengths, the configuration illustrated in this section refers to the
best pattern found by the results. In ACO approach ants are considered as
packets that travel the network changing routing information in order to find
the best path towards the anthill, in this case towards the sink-node. The ant
packet structure is composed by:
All the nodes that belong to the ant path are responsible to update the packet with the proper data, and should update also their internal information regarding the neighbor devices and pheromone levels always taking in consideration the direction of the s ant packet. As a normal protocol AntSense also requires a Hello type packets, in order for the nodes know which are the sensor nodes that are contactable with only one hop. The following structure were used in the hello packet:
It is important to emphasize that the node energy of the
source is also exchange, this procedure allows a to increase the update rate in
order to help ants to decide the best route based in available energy.
The AntSense file structure is divided in three different files (Figure ): antsense.cc, queue.cc, and table.cc. The first one
is the main file, where the routing protocol is written, functions like ant
receive and ant forwards are describe here. This main file requires the
functions described in the last two files, it uses the queue.cc to manage the
data packet queuing, and table.cc to decide the best neighbor device to forward
ant packets and data packets.
Figure 1 – AntSense
structure
All the main and “hard” equations are described in the
table.cc, where a routing table is kept and updated every time a ant packet is
received.
The AntSense model in a first phase was build to behave as a normal ACO
approach, however as the integration in WSNs was being designed the final
solution reflects several aspects of the WSN integration, i.e. reduce the ant
memory to three slots, which allows us to reduce communication time
See the README and INSTALL
files for complete documentation.
.
ASSENT: Ant-based System for Sensor Energy-Efficient Networks
[uk]
Camilo, T., Carreto, C., Silva, J. S. and Boavida, F.,
ANTS2006 - Fifth International Workshop on Ant Colony Optimization and Swarm
Intelligence, Brussels, Bélgica, September 2006.
comments and questions: tandre at dei.uc.pt