A fully distributed node allocation scheme with partition protection for Mobile Ad Hoc Networks

Abstract

In this paper, a fully distributed node allocation scheme, namely Distributed Allocation Scheme (DisAS), is proposed for the purpose of partition protection in Mobile Ad Hoc Networks (MANETs). The proposed scheme controls the locations of nodes in a MANET in a dynamic way. A theoretical upper bound on the probability of partitioning is derived in the scheme. An implementation of DisAS using normal distribution, which is referred to as Normally DisAS ($\mathcal{N}$-DisAS), is also discussed as a case study in this paper. Extensive simulations are carried out to show the effectiveness of $\mathcal{N}$-DisAS in protecting MANETs from partitioning.

Introduction

A Mobile Ad Hoc Network (MANET), as described in [1], is a kind of mobile wireless network formed by a collection of mobile hosts connected through wireless channels. The hosts of a MANET are usually called nodes while direct connections between the nodes are referred to as links. Nodes in a MANET function as network terminals as well as routers. They exchange packets with other nodes and forward packets as intermediate routers at the same time. The routing path of a packet in a MANET is formed by a series of mobile nodes. Packets are forwarded hop-by-hop.

While mobility of nodes in a MANET enables the network to span over a large area, it also causes a highly dynamic network topology, which is a major challenge in the applications of MANETs. In particular, a MANET is connected by mobile devices without centralized connectivity coordination. Therefore, if two nodes in a MANET are not directly connected, the connection between them would depend on other intermediate nodes. A broken link may destroy the entire path. This possibility of link breakage may split nodes into a number of components among which there exists no paths that interconnect them. It in turn results in the inability to transmit packets from the source node in one component, to a destination node which resides in another component of the network. We refer this phenomenon as network partitioning. This problem makes a critical strike on Ad Hoc routing because most routing protocols typically assume that the network is always connected. To enhance the reliability of MANETs, network partitioning should be prevented. The probability of network partitioning needs to be minimized or eliminated. One way is to control the speed, directions or mobility patterns of the nodes. Some of the existing works in this direction include [2], [3], [4].

In this paper, we propose a scheme to evaluate suitable locations for node placement and move nodes toward such positions in a fully distributed manner. We refer to this scheme as the Distributed Allocation Scheme (DisAS). Our proposed scheme uses probability distribution to control the positions of the nodes in the MANET. A particular implementation of DisAS that is based on Normal Distribution is discussed in this paper. This scheme can minimize the probability of partitioning in MANETs. We refer to this scheme as the Normally($\mathcal{N}$)-Distributed Allocation Scheme ($\mathcal{N}$-DisAS). Our proposed scheme is able to maintain good connectivity while keeping satisfactory coverage for exploration and surveillance purposes of MANETs. Triangular geometry and probabilistic principles are used to prove that the network connectivity can be controlled effectively when $\mathcal{N}$-DisAS is used. The effectiveness of this scheme is verified through empirical studies. Our proposed scheme is also able to co-operate with any existing communication protocols and, consequently, is applicable to many real life applications.

The remaining parts of this paper are organized as following: Section 2 introduces the application scenarios of DisAS; Section 3 gives an overview of existing allocation schemes; Section 4 describes how DisAS works and how the probability of partitioning can be estimated; $\mathcal{N}$-DisAS is discussed in Section 5; Section 6 shows the results from our extensive simulations, while the plan of future work and the conclusion are given in Section 7.