When almost everyone began to be besieged with news over how the Internet of Things would change their lives, it was promoted as something brand new, never attempted before: the ability to connect anything that could benefit from being connected wirelessly to the Internet. However, something similar to IoT was first deployed nearly 50 years ago, and since then these Wireless Sensor Networks (WSNs) have expanded beyond their military heritage.
So, even though IoT may be one of today’s Big Things, WSNs have been enhanced by massive improvements in technology, and thousands are still in service today. They’re deployed in markets ranging from agriculture to healthcare, air pollution monitoring, forest fire detection, greenhouse monitoring, landslide detection, medical systems, and even characterizing amusement rides. The question now is how these networks can be integrated into IoT.
A Short History of the WSN
Like so many other technologies that have ultimately resulted in commercial applications, the WSN concept was first developed by the Department of Defense. The U.S. Navy’s Sound Surveillance System (SOSUS) was a passive network of acoustic sensors (hydrophones) on the ocean floor deployed at strategic locations to detect and track Soviet submarines. It was significantly expanded, and although its acronym has changed, the concept (greatly advanced) is used throughout the world’s oceans today.
In the late 1970s, the Advanced Research Projects Agency (ARPA, now DARPA) was exploring new ways to use its new ARPANET packet-switching network whose deployment was rapidly expanding. One of the potential applications was using ARPANET as the communications link for a network of many widely distributed, low-cost sensors, this time above ground, with each sensor collaborating with others while still operating autonomously. Data would be routed to whatever sensor could best make use of it. Thus, the agency’s Distributed Sensor Network (DSN) program was born.
The components included acoustic sensors, high-level communication protocols, a common application platform, processors, algorithms, and distributed software. As DARPA was sponsoring early efforts in AI, it was also included for its potential ability to evaluate signals and or problem solving. To demonstrate the concept, a workshop was conducted in 1978 with distributed acoustic tracking as its target application. There were many contributors to this effort, such as Carnegie Mellon University, Massachusetts Institute of Technology, and the University of Massachusetts at Amherst.
The indoor testbed included signal sources, acoustic sensors, and Digital Equipment Corp. (DEC) VAX computers connected via Ethernet. The success of this limited demonstration did not go unnoticed by DoD and the potential uses for sensor networks were quickly recognized. There seemed to be no limit to the types of sensors used, from the original acoustic types to those monitoring temperatures, humidity, proximity to other objects, and many more.
However, it was apparent that if a network were to use sensors deployed over a wide area, wireless communications would be required, and one of the first applications to use it was acoustic tracking of low-flying aircraft. A prototype used arrays of nine microphones in three concentric triangles with processing performed by a DEC PDP-11 minicomputer and an array processor. The tracking computer was based on three Motorola MC68000 processors, and in addition to Ethernet, the system used microwave radios.
As years passed and DoD focused on network-centric warfare, semiconductor technology advanced dramatically. The small sensors originally envisioned by DARPA started to become available along with more advanced wireless components and protocols. In addition to military applications, WSNs became very popular in a broad range of applications such as infrastructure security, environmental monitoring, industrial sensing and control, traffic control, and many others.
The WSN Defined
The textbook (or rather Wikipedia) definition of a WSN is “a group of spatially dispersed and dedicated sensors for monitoring and recording the physical conditions of the environment and organizing the collected data at a central location (Figure 1). WSNs measure environmental conditions like temperature, sound, pollution levels, humidity, wind, and so on.” As some of the first applications for WSNs were (and continue to be) for monitoring physical or environmental conditions, they have remained part of the definition over the years even though a WSN can be used in many other applications as well.
If a WSN sounds a lot like IoT, it’s because their missions are similar, although how they are deployed is different, as IoT is far more comprehensive in its capabilities and usefulness in the future. That said, as IoT continues to gobble up more and more applications under its broad umbrella, research is still being conducted into WSNs as they are still relevant and have some unique benefits.
A WSN is basically a bidirectional, ad hoc wireless network that does not require infrastructure such as access points that are used in a managed network like IoT. In a centralized approach, each node (typically a sensor) performs routing by communicating with other nodes and forwarding data from one to another, and network formation is controlled by a single device. In a distributed WSN, nodes are autonomous and communicate only between neighboring nodes. There are pros and cons to each approach (Table 1).
The communications protocols used by WSNs typically fall within the IEEE 802.15.4 standards that include ZigBee and WirelessHART. However, some equipment manufacturers use their own proprietary techniques to provide specific benefits in range or other performance metrics. All typically operate in bands designated for unlicensed use. A WSN can be configured in a basic star (point-to-multipoint) topology as well as a mesh network, or hybrid star-mesh network (Figure 2), the latter two providing significant benefits in scalability and elimination of a single point of failure.
While conceptually simple, a WSN can become rather complex when thousands of nodes are deployed. That said, a WSN benefits from its comparative simplicity compared to a managed infrastructure allows it to be configured and deployed quickly. This makes a WSN well suited for emergencies like natural disasters and military systems in which time to deployment is critical. When formed as a mobile ad hoc network (MANET) that is self-configuring and self-organizing, the WSN is appealing because it can move independently in any direction and will therefore change its links to other devices frequently.
Not Just Sensors Anymore
The WSN in its earliest form was basically a group of sensors with wireless communications capability and perhaps the ability to operate bi-directionally, transmitting and receiving data. However, as WSNs have expanded their capabilities over the years, all of their main components have been improved. Sensors, once constrained by their limited processing performance, now include high-performance microcontrollers and an Arm Cortex processor, can operate much longer through energy harvesting or rechargeable batteries, their measurements are more precise, and firmware and reconfiguration can be performed over the air.
A single gateway can handle hundreds of sensors, and they can be supported by dedicated software that connects the system to the cloud for analysis of massive amounts of data. In addition, a wide variety of WSN variants have been developed to support specific applications as well as to increase the performance and applicability of WSNs in general. In short, the line of demarcation between a WSN and IoT has become blurred.
For example, the basic definition of a WSN assumes there is no direct sensor access to the Internet, but this no longer applies to the latest systems. For that matter, IoT sensors don’t generally send data directly to the Internet either, instead using a gateway to aggregate data from the sensors and converting it into a single format, after which it is sent to the Internet via a cellular network or a Low-Power Wireless Area Network such as 6LoWPAN (Figure 3).
IoT gateways are quickly being transformed into edge computers positioned locally where machine learning can reduce the amount of data sent to the cloud, reducing latency to near the vanishing point and retaining proprietary intellectual property at the edge of the network. What this means is that while a high-performance IoT network is far more comprehensive than a WSN, the latter need not be considered obsolete as its sensors can be integrated within an IoT network that ingests data from an entire WSN (or WSNs) to complement its own sensors.
Integrating a WSN into an IoT network can be performed in different ways:
Front-end solution: The IoT’s Internet host is not connected directly to the sensors so the WSN can retain its protocols while still operating within an IoT network using a base station to deliver the data to an IoT gateway.
Gateway solution: In this case, the base station translates the lower level TCP/IP and proprietary protocols and routes the data from one place to another. This allows Internet hosts and sensor nodes to exchange information without a direct connection while the WSN remains independent from the Internet. Sensors can provide Web interfaces to external entities while maintaining their own lower level protocols as in the front-end approach.
TCP/IP solution: Sensors implement the TCP/IP stack such as 6LoWPAN or ZigBee and become functional elements of the Internet. Any Internet host can open a direct connection with them, and the effect is to integrate the WSN with IoT. However, WSN sensors can no longer use their native protocols.
Hybrid solution: The nodes within the WSN can access the Internet directly and can be mapped to base stations that have Internet access as all sensor data must pass through them to access the central system.
Access point solution: WSNs become trees with multiple roots, with leaves viewed as standard sensor nodes and all other elements of the tree as Internet-enabled nodes. All sensor nodes can access the Internet in a single hop. Using this approach, the nodes have more resources and can implement faster wireless protocols.
Any of these approaches and other variants have their own advantages and disadvantages, but regardless of which one is chosen, the result is the integration of the WSN within IoT.
The WSN has a long, impressive history of providing benefits using widely dispersed sensors that communicate with each other to provide information useful or even vital for a specific application. As there are so many of these networks in service today, it would make little sense to simply abandon them, not only because WSNs have advanced so far, but also because they can be used as a subset of IoT to complement and in some cases greatly expand the network’s resources.
The question, of course, becomes how to do this with the least disruption of the WSN and its native protocols and general method of operation. Fortunately, although this is not a simple task, the flexibility of IoT has the potential to make this possible in the vast majority of cases. Nevertheless, it is necessary to discuss whether full integration at the network level using direct TCP/IP connections is required for every application.
The general consensus seems to be that for some applications, such as Supervisory Control and Data Acquisition (SCADA) systems, the difficulty may not be economically worth the benefits, especially considering the truly enormous number of SCADA systems and services throughout the world in utilities, energy production, and many other mission-critical applications.
It should also be noted that security, which for the most part was not fully implemented in WSNs, is now inherent in most systems. If this were not the case, providing the same level of security as embodied in IoT would likely be much more difficult. Nevertheless, security is just one issue that must be considered when integrating a WSN within IoT, and as seems likely, this and all other concerns will be addressed sooner rather than later.