Introduction

Abstract

Environmental monitoring has generated a large amount of information for scientists. The purpose of this experience is to improve our knowledge about the impact on nature of human being. Data are collected by sensors, which are usually measuring temperature, humidity or air pressure. However, any sort of gas may be monitored for the need of a specific project. Fifteen years ago, the only suitable way to centralize information coming from sensors was the use of cables. This was an important barrier to the development of sensor networks. Nowadays, Wireless Sensor Networks have permitted new way of research. The nodes are organized on an ad-hoc wireless architecture (no fixed infrastructure is employed). Sensor networks often have a root point called sink where data are collected. The radio coverage of each node is usually not large enough to reach a sink. Hence, the nodes have implemented a protocol called Multi-hop routing to allow them to send information to the sink via other nodes. The large size of our sensor network forces us to employ multiple sinks.
This document describes and discusses an architecture for implementing a solution to centralize data from the multiple sink. We proposed a solution and practical tools which collects the sensor measurements and dispatches them through a TCP/IP-based network. The node sink is connected to an embedded computer via the RS-232 port where the data are temporary stored inside a database. Then, the data saved in every sinks are sent to the centralized database server. In the first part of this article, we described the constraint for the Wireless Sensor Network. The second part is a description of the architecture of how the data are collected. Then, a rapid description of the way to represent the data collected via interactive maps. The fourth part is a description of the sensor integration sensor on the tinynode. And final, the last part summarizes the existing works related to our.

Wireless Sensor Network

The sensor will generate periodically information that have to be computed by a computer. Therefore, the usual way to set up a sensor network is to employ wireless communication nodes such as Tinynode or MICA. These nodes are ultra-low consumption which allows them to run several years with a single battery if the amount of data transmitted is low. The main advantage of this architecture type is its flexibility. New sensors can be added without any need to change the infrastructure.
The Wireless Sensor Network is composed of multiple nodes connected to each other with a root called Sink. Most of the time, many nodes are out of the radio coverage range of the Sink as described in Figure 1. The sensor node 3 can not reach 1 directly because it radio coverage is not large enough. However, it is able to communicate with node 2 which can communicate with number 1. Therefore, the communication between 1 and 3 has to transit through number 2. In this example, the information is transferred via one node but in other situation it may have more than one transitional node.

Cluster-Based Routing

The final purpose of these protocols is to bring the information coming all over the network to a simple node called sink. Then, the sink node sends the information to another interface (RS-232 for tinyos). The Figure 2 represents an Ad-hoc network with one central point (sink). Any message emitted from any nodes on the ad-hoc network will be transport through the multi-hop protocol to the sink. It may transit via several nodes. It is important to understand that a node only needs to know the address of its parent and nothing more.

Usually these types of protocol are implemented to work with only one sink in the whole network. However, if the number of node is becoming significantly large, the node directly connected to the Sink will rapidly run out of battery. This is due to the fact that they will transit packet coming from all the ad-hoc network. Therefore, a multi-sink router protocol has to be used.  

 Data collection timing

For pollution monitoring, the sensor should emit periodically the data measured. The time between two sent depend on many factors as: how quite can change the value measured, the battery life expectation, and the amount of data which the architecture can managed. To save battery life time, the data may be aggregated by nodes. Then, only a packet containing several measurements will be sent.
There is three different timing models:

• Periodic simple aggregation: simply aggregating data during a certain amount of time and after sending them together.
• Periodic per-hop: similar to periodic simple aggregation but send the aggregated data as soon of all its children have sent data.
• Periodic per-hop adjusted: Adapt the frequency of aggregation sent with their position on the multi-hop tree.

This solution may look interesting but they have the disadvantage to make the system less real-time because the data may be trapped during a long time before being sent.

Broadcast

In an ad-hoc network, broadcasts may be a serious issue. The number of node in the network may vary unpredictably, because they do not have any fixed infrastructure. For our project, the problem is quite simpler because nodes are not able to spontaneously generate a broadcast message. The broadcast messages will only be set up by the sinks and are used for sending instruction to the nodes. An essential characteristic for broadcasting is the ability to identify two different broadcast messages; therefore a sequence number has to be added to its header.
The principle of broadcasting in an ad-hoc network is called flooding. Each node receiving a broadcast message will automatically retransmit it. However, there are a few inconveniences with this technique : redundant broadcast messages, collision between broadcast messages, starving nodes .