Wireless sensor networks (WSNs) are emerging technologies which is composed of several powerful base stations and hundred or thousand of small and autonomous device called sensor nodes. In WSNs, the deployed sensor nodes perform distributed observation locally and transmit their sense readings back to the base station over a tree structure. The sensor nodes are usually densely deployed to unattended environment to observe physical phenomenon such as temperature, humidity, light, precipitation, seismic intensity, and etc.
A sense node is resource-constrained with low computation power, limited storage, poor communication, low power capacity, and etc. Operated by battery makes power consumption an important issue for sensor nodes. The failure of a set of sensor nodes by energy depletion can lead to a partition of the sensor network and loss of important information for applications.
There has been much research on the topic of energy-aware routing protocols in sensor networks [16], [17], [18], [19], [20], [21], and efficient data processing approaches [13], [14], [15] for reducing power consumption of sensor nodes and further increase the lifetime of the whole network.
1.1 Requirements
In this paper, we design a protocol which focuses on efficient data processing for reducing data volume rather than energy-efficient routing protocol. To design this protocol in wireless sensor networks may face some challenges:
z Low computation overhead
In wireless sensor networks, all sensor nodes are resource-constrained devices. An energy-efficient protocol should try to minimize the computation overhead and save the maximal total energy in the network. Applying common data compression techniques to reduce data volume cost too much overhead for sensor nodes. Sensor nodes which
perform encoding and decoding in these data compression techniques may drain their power quickly due to the complex computation. We give a discussion in section 2.3 to show that common data techniques are not suitable for wireless sensor networks.
Therefore, designing an energy-efficient protocol should keep computation overhead in mind.
z No data distortion
In some efficient-aware protocols [17], [21], [22], their schemes have some level of data distortion. Data distortion provides imprecise data for applications. These protocols with data distortion limited themselves to applications which need precise sensed data.
z Fault Tolerance
One of the characteristics in wireless sensor network is unreliable links. Packet loss occurs more frequently than wired network due to limited bandwidth and interference. A mechanism is required for recovering the lost packet. Designing an energy-efficient protocol should take car of this factor.
1.2 Cluster-based wireless sensor networks
In wireless sensor networks, the resource constraints on sensor nodes make it an important challenge to develop efficient data processing techniques. It is inefficient to directly propagate the raw data which sensor nodes observed to the base station. Instead, raw data should be processed, aggregated locally, and reported back to the base station to avoid energy depletion. In recent research, a cluster-based network has been discussed in the literature to concretize this idea. In this architecture, the network is divided into small clusters and each cluster has one sensor node acting as aggregator which collects all sense readings of sensor nodes in the cluster and performs intra-network data aggregation. The sensor nodes which are the members of the cluster transmit their raw data to their own aggregator. Only the sensor nodes which are aggregators report their aggregated data to the base station. Therefore, the
clustered-based networks can conserve energy depletion of the sensor nodes because that only aggregators need to propagate their collected data to the base station while other sensor nodes just transmit their raw data to their aggregator. Moreover, the aggregator nodes fuse the raw data into one single data by intra-network aggregation to reduce data volume and also benefit the power consumption for the whole networks. Note that if the aggregator node is chosen a priori and fixed throughout the system lifetime, it is clear that unlucky sensor node chosen to be the aggregator would die quickly. Thus the aggregator nodes should be rotated after a period of time in order not to drain energy of single sensor node.
1.3 Related Work
There have been many proposed energy-efficient schemes for wireless sensor networks in the literature. These schemes can be roughly classified into two categories. The first one is efficient-aware routing protocols, which can be further divided into three main categories, data-centric, hierarchical and location-based [3]. The other one category of energy-efficient schemes is efficient data processing which reduces the transmitted data through the networks to further reduce power consumption.
1.3.1 Energy-aware routing protocol
Directed diffusion [18] is a data-centric scheme using a naming scheme where data generated by a sensor node is named by attribute-value pairs. An interest is defined using a list of attribute-value pairs such as type of object, interval, duration, location, and etc. The interest is broadcast by a sink and each node receiving the interest performs caching for later use. The interest entry contains several gradients which are reply links to neighbors from which the interest was received. By using interest and gradients, directed diffusion enables sink and nodes to establish empirically good paths between them to achieve power saving.
LEACH [6] is one of the most popular hierarchical routing protocols for wireless sensor
networks. In LEACH, a sensor network is divided into several clusters. A cluster contains several nodes and one cluster head. In this architecture, local cluster heads act as routers to route data from the members of the clusters to base stations. It will save energy because that data propagation is only performed by cluster heads and all other nodes transmit their data to their own cluster heads which are only one-hop away from them. There have been several research based on LEACH to further improve the performance of power saving.
Lindsey et al. proposed a power-efficient data gathering scheme for wireless sensor networks [19]. The basic idea is that the nodes in the networks are organized to form a chain to the base station. The chain is constructed by the sensor nodes themselves using a greedy algorithm staring from some nodes. After the chain is constructed, each node on the chain receives data from its last neighbor, aggregate with its own data, and then transmits the fused data to its next neighbor on the chain. Although it reduces power consumption by decreasing the number of transmissions and reception by using data aggregation, knowing the topology for each node can introduce significant overhead especially for dense networks.
1.3.2 Efficient data processing
Other energy-efficient schemes are focus on data processing which reduce transmitted data for power saving. Some approaches proposed novel data compression schemes rather than common compression techniques to reduce data size due to heavy overhead of these common compression techniques.
Chou and Petrovic et al. [14] proposed an adaptive signal processing approach to reduce power consumption in wireless sensor networks. The base station constructs a tree-based codebook for compression with side information and broadcast to all sensor nodes in the networks. Each sensor node compresses its observation according to the given codebook to reduce data size, and transmits to the base station which then decodes these compressed data with side information. There are some drawbacks in their scheme. The base station needs to
continuously track and exploit existing correlations in sensor data for decoding with side information. Moreover, it is possible for their scheme to make a decoding error due to the correlation noise. As a result, it would have some level of data distortion in their scheme.
PINCO [15] presented a power saving scheme by reducing redundancy in the data collected by sensor nodes. Each node receives the measurement from its neighbor and performs data fusion. The prefix of observations which the same are combined together to reduce the transmitted data size. TiNa [22] utilized temporal coherency to reduce the amount of data transmitted by all nodes. Their scheme define a user-specified parameter, called tct, which specifies the degree of which the user tolerant to the change of sense readings. The larger tct, the greater degree of distortion of sense readings would be. Similar approach is also proposed in [21] which utilized two thresholds, hard threshold and soft threshold. Only the nodes whose sense readings are beyond the hard threshold and equal or greater than soft threshold transmit their sense readings. However, this approach is not good for applications which need to periodically report the sense reading of sensor nodes. Applications may receive no data because no data reaches the thresholds.
1.4 Contribution
In this paper, we proposed a data-efficient protocol in the wireless sensor networks to reduce transmitted data volume as well as providing the same level of quality of sense readings. Each sensor node in our proposed protocol can save energy efficiently and further increasing lifetime of the whole network. Moreover, no data distortion will occur in our proposed protocol so that it is independent to upper applications in wireless sensor networks.
Our proposed protocol is easy to implement and suitable for any cluster-based wireless sensor networks which perform intra-network data aggregation.
1.5 Synopsis
The rest of the paper is structured as follows. In the next chapter, we introduce the preliminaries needed in our proposed protocol. Chapter 3 describes the environment which our protocol discussed and gives a brief description of our proposed protocol which consists of three phases. In chapter 4, we analyze our protocol and give simulations to show that our performance on power saving. Finally, we conclude our proposed protocol in chapter 5.