The use of Wireless Sensor Networks (WSN) in a variety of (Industrial) Internet of Things scenarios has gained popularity over the past years. Whereas WSN are known for their limited computation and communication resources as well as limited battery power, the right IoT data aggregation methods can make all the difference in your IoT strategy.
A standard definition of a Wireless Sensor Network describes WSN as follows:
“A WSN is an ad hoc network consisting of a set of sensor nodes, randomly fixed or dispersed in a given geographical area, communicating via a wireless link to autonomously collect, process, and transmit data in their environment to a special node, considered the collection point, called a sink node.”
Even though WSN has a vast number of applications including healthcare, security, and military applications, the network also faces a variety of challenges. These include energy consumption and fault tolerance. These challenges, in turn, feed into another typical hurdle for IIoT: the sheer growth of the number of connected IoT devices and their applications, as well as the large amounts of data generated by these devices.
In such scenarios, we have highly heterogeneous traffic loads and a lot of redundancy in the transmission of data. Another hurdle is the fact that IoT sensor nodes only have a limited capacity to process data. And processing consumes battery life. In scenarios when we deal with reduced battery power, the sensor network becomes more susceptible to failure. So how do we overcome all of this? Specifically, what techniques are there in place to support this effort in setting the ground for a well-functioning IoT analytics layer?
Challenges to the implementation of wireless sensor networks in IoT scenarios
To improve our use of this complex technology, we need a closer look at some of the typical WSN challenges. These include:
- Data volumes: This calls for re-conceptualizing of the very way things will be connected with one another in favor of ways that reduce the generation of massive volumes of raw IoT data as well as facing the challenge of processing and storing the data from connected IoT devices;
- Data heterogeneity: Facing millions of different connected and interconnected IoT devices, one is confronted with the need to organize the incoming data and reduce its complexity;
- Highly dynamic IoT landscapes: Facing constantly changing conditions such as night, day, working hours, disconnecting and reconnecting IoT devices, as well as highly heterogeneous device landscapes of ever-growing numbers of IoT devices connected to a network, there arises the need for managing IoT ecosystems in a systematic way.
In view of these challenges to the use of WSN in IoT applications, different IoT data aggregation methods can overcome both energy-efficiency and data transmission hurdles. The most common definition of data aggregation is the process of fusing the data from multiple sensors to minimize redundant transmission. In this way, only the fused information is provided to the base station. Typically, data aggregation fuses data from multiple sensors at intermediate nodes and transmits the aggregated data to the base station.
Why data aggregation
According to a recent definition, data aggregation is the process of gathering and summarizing data from multiple sources. Aggregated data is normally found in a data warehouse. There, it can give answers to analytical questions and greatly reduce the time required to query large data sets.
The main rationale behind data aggregation is that it minimizes energy depletion and the required network bandwidth. The use of different IoT data aggregation methods eliminates redundant data. This reduces network traffic by significantly minimizing the number of sent data packages. IoT sensor nodes can also eliminate redundancies in the data received from neighboring nodes before transferring the final data packages.
Another aspect to consider is the tradeoff between bandwidth and distance. For example, with Sigfox or LoRa, you can only send 2 bytes every 10 minutes but over very long distances.
Questions of sustainability are also part of the discussion. Since sensor nodes are powered by batteries, saving energy and extending battery life is essential to the IoT data collection effort. Data aggregation is considered an energy-aware data collection technique and is preferred in scenarios where extending battery life is crucial. It is even known to increase the lifespan of WSNs.
Energy-aware data aggregation methods include clustered aggregation, tree-based aggregation, in-network aggregation, as well as centralized data aggregation that specifically considers the energy consumption of sensor nodes.
Data aggregation methods in IoT sensor network settings
All things being equal, there is a need to identify suitable data aggregation techniques to collect and analyze incoming data. Typically, at this level, we differentiate between flat IoT data aggregation methods and a hierarchical approach to data aggregation.
In flat wireless sensor networks, all sensors play an equal role—there is no hierarchical arrangement. Every sensor node serves the same purpose and all IoT sensor nodes are peers. One disadvantage of flat wireless sensor networks is that data aggregation takes place only in the sink node area. As a result, network delay can be high. Also, if the sink node fails, this negatively impacts the entire network.
With the hierarchical approach to wireless sensor networks, there is a hierarchy among the individual nodes based on their capabilities. Roughly, these are divided into base stations, cluster heads, and sensor nodes. The sensor nodes within a given cluster communicate with each other and then communicate with the cluster head. More computing power and increased network transmission capabilities mean less battery life. So one of the main goals of this routing method is to achieve better energy efficiency for the sensors within a cluster.
This is a hierarchical method best suited for large-scale energy-constrained sensor environments. In such scenarios, it is not efficient for the sensors to transmit the IoT data directly to the sink node (base station). Rather, sensors transmit data to a local aggregator, also known as a cluster head. The cluster head aggregates data from all the sensors in its cluster and transmits it to the sink node. The cluster heads can communicate with the sink node directly via long-range transmissions.
They can also do multi-hopping through other cluster heads. The typical protocols here include clustered diffusion with dynamic data aggregation (CLUDDA), Low Energy Adaptive Clustering Hierarchy (LEACH), and Hybrid Energy-Efficient Distributed Clustering Approach (HEED).
Chain-based aggregation or string-based aggregation
In some scenarios, if the cluster head is located too far from the sensors, communication between the sensors and the cluster head might consume excessive amounts of energy. On such occasions, it is more efficient for sensors to transmit data only to their closest neighbors in the network. Chain-based data aggregation is a hierarchical method whereby each sensor transmits only to its closest neighbor. Nodes are mostly organized into a linear data aggregation chain.
The node that is located farthest from the base station initiates chain formation. At each step, a node’s closest neighbor is selected as its successor in the chain. So a node receives data from one of its neighbors and merges the received data with its own data. It then transmits the fused data further down the chain to its next neighbor. A lead node, similar to the cluster head in cluster-based aggregation, transmits the aggregated data to the base station.
An example of a chain-based data aggregation protocol is the so-called power-efficient data gathering protocol for sensor information systems (PEGASIS).
In this scenario, data is aggregated through the creation of a data aggregation tree. Sensor nodes are organized in such a way that data aggregation takes place at intermediate nodes along the “tree”. The so-called “root node” only receives an already structured representation of the data. This aggregation technique is suited for applications that necessitate in-network data aggregation. One of the main challenges of tree-based aggregation is the creation of an energy-efficient data aggregation tree that optimizes the network lifespan and minimizes the number of transmissions.
On average, tree-based methods are known to have high overhead, high energy uniformity, as well as greater strength, flexibility, and scalability as compared to cluster-based methods.
This method is based on dividing the region of a sensor network into several grids. A set of sensors act as data aggregators in pre-defined regions of the sensor network. So we have a data aggregator (also known as an integrator) that is fixed in each grid. And the array of sensors acts as an aggregator/integrator within this particular region of the sensor network. The sensors in that particular grid transmit the data directly to the data aggregator that aggregates the data from all IoT sensors within the grid.
In grid-based aggregation, the individual IoT sensors within a grid do not communicate with each other. Grid-based data aggregation is known to adapt to dynamic changes in the network.
Structureless data aggregation does not involve any kind of architecture. Communication takes place from any node to any node within the network. In some cases, as for example in event-based applications that vary by event region, structureless aggregation is the preferred approach.
While building your IoT strategy, you want to be able to flexibly adjust your network both to new requirements and to a growing number of connected devices that need to be kept in check. An IoT development platform helps you implement the method that suits you best, develop applications, rapidly test in a secure environment and roll out your apps over the air on any number of IoT devices and device groups.
You monitor the status of all your IoT devices from within a single venue, getting instant feedback from your devices. This is how you find out what method works best for you, striking the right balance between energy efficiency, performance, and data accuracy. Get in touch for an in-depth discussion.
About Record Evolution
We are a Data Science & IoT team based in Frankfurt am Main, committed to helping companies of all sizes to innovate at scale. So we’ve built an easy-to-use development platform enabling everyone to benefit from the powers of IoT and AI.