Swarm intelligence and its applications to wireless ad hoc and sensor networks.

Swarm intelligence, as inspired by natural biological swarms, has numerous powerful
properties for distributed problem solving in complex real world applications such
as optimisation and control. Swarm intelligence properties can be found in natural
systems such as ants, bees and birds, whereby the collective behaviour of unsophisticated
agents interact locally with their environment to explore collective problem solving
without centralised control. Recent advances in wireless communication and digital
electronics have instigated important changes in distributed computing. Pervasive
computing environments have emerged, such as large scale communication networks
and wireless ad hoc and sensor networks that are extremely dynamic and unreliable.
The network management and control must be based on distributed principles where
centralised approaches may not be suitable for exploiting the enormous potential of
these environments. In this thesis, we focus on applying swarm intelligence to the
wireless ad hoc and sensor networks optimisation and control problems.
Firstly, an analysis of the recently proposed particle swarm optimisation, which is
based on the swarm intelligence techniques, is presented. Previous stability analysis
of the particle swarm optimisation was restricted to the assumption that all of the
parameters are non random since the theoretical analysis with the random parameters
is difficult. We analyse the stability of the particle dynamics without these restrictive
assumptions using Lyapunov stability and passive systems concepts. The particle
swarm optimisation is then used to solve the sink node placement problem in sensor
networks.
Secondly, swarm intelligence based routing methods for mobile ad hoc networks
are investigated. Two protocols have been proposed based on the foraging behaviour
of biological ants and implemented in the NS2 network simulator. The first protocol
allows each node in the network to choose the next node for packets to be
forwarded on the basis of mobility influenced routing table. Since mobility is one of
the most important factors for route changes in mobile ad hoc networks, the mobility
of the neighbour node using HELLO packets is predicted and then translated into a
pheromone decay as found in natural biological systems. The second protocol uses
the same mechanism as the first, but instead of mobility the neighbour node remaining
energy level and its drain rate are used. The thesis clearly shows that swarm
intelligence methods have a very useful role to play in the management and control
iv
problems associated with wireless ad hoc and sensor networks. This thesis has given
a number of example applications and has demonstrated its usefulness in improving
performance over other existing methods.