The wireless sensor network is fundamentally a distributed network of constrained devices. There are a host of considerations that must be taken into account when planning and implementing a wireless sensor network – considerations that are often overlooked in the lavish space in which traditional IT operates with nearly extravagant disk space, CPU and working memory. In contrast, the wireless sensor is a targeted device with limited memory, limited power, and limited processing capabilities, which is then deployed in the harshest environments.
The first consideration is the deployment of the devices themselves. Depending on the coverage extent, a wireless sensor network may cover a large geographic area. It may be hard to believe, but projects often do not have an unlimited amount of time in which to deploy; for that reason, it may well be necessary to employ several field crews tasked with deploying the sensors in the field in parallel.
Any project planning must take into account the oversight of this distributed workforce, in order to ensure consistent efficiency and quality. Running distributed teams is a whole topic in and of itself, but it’s valuable just to be aware of the challenge that such a distributed and geographically dispersed team might present.
A second consideration in planning a wireless sensor network deployment is with respect to the communications network. In traditional IT applications, the network is ubiquitous; a commodity that can be assumed as readily as power. In the field, neither power nor communications can be assumed. One of the first decisions is whether to implement a communications backbone or to rely on commercially available access such as cellular, WiMax or something similar. The consideration will likely depend on the tolerance to failure – for mission critical applications, the service level of commercial communications may not be acceptable.
Of course, when making the choice to make this investment, keep in mind that the sensors themselves are often of a limited lifetime and incorporate that factor into the cost/benefit analysis. Cellular can often provide good coverage at a reasonable cost, but depending on how the network is deployed, cellular coverage can be spotty. For situational awareness applications, cellular can also be least reliable when it is most needed – during the situation. Cell towers are typically fine on a blue sky day but can go down in a storm.
Another challenge may be software development, particularly with respect to distributed algorithms and data routing. Although the landscape is changing with the increasing penetration of multi-core CPUs, many developers do not have knowledge or experience developing distributed or parallel applications. There are many factors in such development that should be considered; for example, how memory is parsed among the distributed processors.
Two common architectures are MIMD (multiple instruction, multiple data) and MISD (multiple instruction, single [shared] data). Wireless sensor networks are becoming more commoditized, and often have the glue architecture already in place, but it is valuable to understand the constraints.
Security can be a significant challenge and a significant risk to the project. There are a variety of security challenges. An attacker node can spoof its identity as a sensor in the network, and gain access to the network. From here, it may be able to gain access to potentially sensitive information (for example, data about the state of a utility network) or to even insert malicious instructions into the stream (for example in a SCADA, or supervisory control and data acquisition, network that controls industrial or utility devices).
An impostor may also be able to compute the location of actual nodes in the network from signal strength and angle of arrival. With this information, an attacker may find and physically compromise the network nodes. Wireless sensor networks are also susceptible to denial of service (DoS) attacks; because the sensor is already a constrained device, placing a significant amount of data into the data stream can cripple the network. Ensuring that sufficient security is planned for is an important component of the deployment.
Finally, testing can be a significant challenge. Because of the constraints uniquely imposed by the field, it may be difficult or even impossible to replicate conditions. Communications quality will vary, network topologies and spatial relationships will differ. Environmental factors will certainly not be identical.
Ensure that sufficient time is allowed not only for testing in development, but in deployment as well, and that sufficient time for addressing these issues is provided. Many sensor deployments have gone well right up until the point that they didn’t, and then there was no room left to rectify. Insufficient testing can kill even a well-executed wireless sensor network deployment.
It is important to ensure that the project manager or project engineer has experience with and knowledge of field technology and wireless sensor network deployments. Deployment in the field is very different than deployment in the office. The good news is that as the industry matures, and as IT/OT convergence continues to happen, these challenges will be better understood and better factored into the project planning process.
