Optimum Node Density in Wireless Sensor Networks

BSC Class Notes » Wireless Sensor Networks & Distributed Systems

Routing Protocols & Power Usage

The paper of technology, Critical Density Thresholds in Distributed Wireless Networks, outlines the challenges and possible solutions to building a wireless sensor network. A sensor network must be able to communicate with all points within itself, but optimizing this communication is a primary concern. This paper considers the following schemes to implementing a wireless sensor network: network connectivity, multi-path reliability, neighbor count, Hamiltonian cycle formation, multiple- clique formation, and probabilistic flooding. All of these aspects are covered in great detail in this paper. Considering all of these, the major goal of this paper is to consider the best design of resource-efficient wireless networks.

The applications considered in this paper are various and notes that there are possible uses for wireless sensor networks that have not even been considered yet. However, wireless sensor networks should be designed in such a way that, even though the uses for the wireless sensor network has not been preconceived, it should still be a reliable source of data transmission. Applications for wireless sensor networks considered are environmental sensing and disaster recovery. These are two very different applications that can be designed in the same way as each other.

The major factor differentiating these two applications, as stated in the paper, is that environmental sensing could use static nodes whereas disaster recovery would be better suited implementing a dynamic topology of wireless sensors. Also, the applications that will be implementing these wireless sensor networks will probably be very large-scale, with hundreds to thousands of wireless communication nodes.

The major focus of this paper is to find the critical density for wireless sensor networks such that all the wireless communication nodes are connected to the entire network, or graph, of nodes. If the transmission power of each node is set too low, then there is a high probability that there will be wireless nodes that are not connected to the graph at a given time. However, if the transmission power of the nodes is set so high that there is an abundance of neighbors to the wireless node, then there may be too much power being used by the wireless nodes, making it inefficient.

This paper also points out that having too many wireless node neighbors can have another adverse effect on the efficiency of the wireless sensor network. If there are too many wireless nodes in a given area at one time, then transmission of data between all of these nodes can cause collisions and data loss because so many wireless nodes are trying to communicate at once. Although the graph is connected, communication across the graph becomes increasingly difficult due to high incidence of data loss. This is another valid point the paper makes for a need to create an efficient wireless sensor network.

There threshold which determines a wireless sensor network to be dense enough to have a high probability of being connected, but at the same time sparse enough such that there is not a great amount of overlap between neighboring wireless sensors. This desired threshold is known as the phase transition of the graph and is the ideal operating point for the system. The authors of this paper measure these phase transitions with random graphs and fixed radius communication distances for the wireless nodes.

The randomness of the graphs is not the only challenge that designers of wireless sensor networks need to overcome. A major factor of wireless sensor networks is that they are dynamic and may often change their topology. This paper addresses the challenges of a dynamic topology and points out that the speed of the nodes in the wireless sensor network plays a critical factor when determining its critical density.

A critical density formula needed to be addressed in order to determine the phase transition of a given wireless sensor network. The paper determines this formula by randomly arranging wireless nodes in an area of unit size. Each wireless node is said to transmit a fixed radio power where the transmission can be received by all other nodes that are within a given radius. This allows nodes to communicate with each other if they are a certain distance apart, but can not maintain communication if the wireless nodes move greater than that distance apart.

After establishing this formula, experimental results could illustrate the existence of phase transitions and critical density thresholds for a number of global properties in wireless networks. An interesting approach to solving the challenge of data collision is to calculate a Hamiltonian cycle for the graph of wireless nodes, which would allow for a token ring to travel through the network. This transmission of data, which would travel through every node, but never to the same node twice, would eliminate the possibility of lost data due to data collision.
When calculating the graph of the wireless nodes, however, the paper points out that this is not a simple random graph model. The nodes will be moving and may lose transmission with nodes and gain transmission with others. This compromises the ability to calculate a Hamiltonian circuit.

The paper concludes with a formula that could possibly be used to calculate an average transmission power range for each wireless node to optimize the network. Other factors, such as environment and speed of wireless node travel, need to be taken into account when designing a wireless sensor network.

In the paper of application, An Analysis of the Optimum Node Density for Ad hoc Mobile Networks, is a discussion of the need for communication over wireless channels through nodes and the ability of these nodes to move throughout space while maintaining connectivity.

An example of stationary wireless networks is given due to its relevance to a wireless sensor network. A mobile wireless sensor network should be built off of the ideas of a stationary sensor network. The point is made, though, that algorithms must be designed for mobile routing protocols.

The major challenge of such a network is to optimize a wireless sensor network in such a way that wireless nodes can communicate with each other without wasting battery life or bandwidth. Arising from this need to create optimum network connectivity are several routing protocols that the paper outlines.

One such routing protocol is used to discovery a route through nodes when one is necessary. There are complex algorithms holding these routes together which are also outlined in this paper. One maintaining factor of these routes is that they are kept loop-free to eliminate the possibility of an infinite loop of data or cross communication and data collision of packets being sent across the network.

Building off of the technology we currently have, the routing protocols implement an idea of the current routing protocols we use for the Internet today. It has been simulated using the OSI seven layer network architecture and models of IP routing and UDP seen in the very common TCP/IP network. Transmission also is initiated with a Request To Send packed and ended with an Acknowledgement packet, also used in current technologies.

These protocols cannot guarantee packet delivery, however. It does pose the best possible network model that can be generated using the wireless sensor nodes' available resources. Packet loss and transmission disconnect are common among wireless sensor networks and must be considered when building a wireless sensor architecture.

In simulations, a major determining factor of successful packet delivery was determined by node velocity. It must be considered, that the faster the wireless nodes are moving, the more ground they will cover and, by probability, the more nodes they will come in contact with over a given period of time. However, if a connection is established with a certain node, then the connection has a greater possibility of being disconnected due to the greater speeds of the nodes. The paper gives several different models with varying approaches to try and deal with this drawback.