Performing Without a Net

The largest single barrier to the wider deployment of wireless devices at the sensor level is that they have no standard network layer protocol. This is not just a typical emerging market issue that can be remedied by a few new standards.

Aw 4786 Ivwireless

Rather, the issue is that the TCP/IP protocol suite is not well-suited for resource and energy constrained devices such as wireless sensors. It is one thing for an emerging market to need one to two new standards, but when a new market cannot easily leverage TCP/IP (transmission control protocol/Internet protocol) and its many associated services and technologies, this represents a major hurdle.

Wireless sensor networks (WSNs) have not used TCP/IP because of the extremely limited resources available on wireless sensor platforms. Instead, sensor networks integrate with TCP/IP using mains-powered gateway devices. Ironically, it was the lack of TCP/IP that attracted many computer science researchers to this field. WSNs offered an intriguing research platform because conventional networking assumptions did not apply. Therefore, fundamentally new laws and protocols might emerge at layer 2 and layer 3 of the International Standards Organization (ISO) model.

Commercial network layers

There are a number of existing and proposed network layers for WSNs. Besides the commercial offerings, the Instrumention, Systems and Automation Society (ISA) SP100.14 working group is looking into standard wireless networking for what SP100 defines as Class 4-5 applications, meaning monitoring-only applications. Some form of networking standard for sensor networks may emerge from this group in 2007. The Institute of Electrical and Electronic Engineers (IEEE) also has a working group that is exploring the implementation of Internet Protocol version 6 (IPV6) on sensor networks. This work targets home appliances that have no energy constraints.

The choice of IEEE radio standards is driven by power constraints, rather than protocol preference. There are two families of IEEE 802 standards that apply to industrial wireless devices at the network edge. First is the familiar IEEE 802.11 family, which underlies the wireless fidelity, or Wi-Fi, technology. The second standard is IEEE 802.15.4 used by many sensor network vendors. The IEEE distinguishes between these two standards based on their transmission range. It by classifies 802.11 networks as a local area networks (LANs) and 802.15 networks as personal area networks (PANs). In the commercialization of Wi-Fi and the ZigBee wireless specification, however, the range is not the major differentiator. Power consumption is really the key difference. IEEE 802.15.4 networks can be designed to operate at very low power levels that can extend battery life beyond that of equivalent Wi-Fi devices.

TCP/IP networks have boomed historically because a single protocol (IP) is used at the Network layer (L3) for all types of services. Thus, IP forms a “narrow waist” in the stack. This allows a single IP network infrastructure to provide services for many different applications, and allows IP to operate over many different links and physical media. Currently no such “narrow waist” exists for wireless sensor networks.

Some academic researchers believe that the Data Link Layer (L2) rather the network layer (L3) may serve an IP-like unifying role for wireless sensors. These researchers have proposed that a new link layer (L2) be used to create a unifying assumption for wireless sensor networks. This would require a “translucent” L2 that enables multiple network (L3) protocols to coexist, and would inform the L3 protocols about lower level considerations such as device discovery, energy management, timing, and the like. At this point, such research is not mature enough to have developed and demonstrated robust sensor network protocols on a wide scale. The suitability of a single “sensor protocol” (SP) in real-world applications is unproven, though its value from a network architecture standpoint is clear. 

Harry Forbes,

 

More in Control