The projected growth of the Internet of Things would have made Carl Sagan proud: “billions and billions” of devices and sensors, capable of communicating via the Internet. It is exciting to consider the benefits of being able to remotely communicate with each of these sensors and get access to important data. Benefits can range from improving operation of elevators or large air conditioning equipment, to ensuring wind turbines or solar farms in a remote area operate safely with no down time.
Merely connecting sensors to create a network only constitutes a nervous system. But are these connected systems “smart”? The moniker “smart” is bandied about so much these days that it is important to identify the true characteristics of a “smart” system. Consider the human body – an amazing example of a system that is able to integrate, process and respond to data from the environment to maintain homeostasis. Using this system as a template, what is needed to make a set of connected devices in a system actually “smart”? A brain, of course.
Surely, many simple systems today are able to perform optimization and control. A thermostat maintaining temperature in a room is one such example. However, when it comes to larger, more complex systems, the challenge becomes integrating and optimizing the performance of multiple systems so that they work in concert. For these multi-layered systems, it is vital that the “brain” use methods to analyze and parse data with the system of systems view.
What, then, are the critical elements that would enable a brain powering a complex system to work well? First, it must have accurate pattern recognition to be able to sort through the inherent noise generated by a multitude of sensors. Second, it must be able to correctly identify anomalous behaviour through establishing a baseline for healthy behavior. Third, it must have an intimate understanding of how the entire system works together, with a full cognizance of cause and effect.
Lastly, to fully realize the promise of a smart system, operation needs to be automatic, autonomous, and reliable – at scale. Leveraging all of the machine-generated dynamic data should not require flocks of specialized “data scientists” staring at pretty data visualizations. Don’t forget that this lofty goal of truly smart IoT systems can only be realized with the help of advanced algorithms, not just a few sensors strung together with bailing wire.