Thanks to AI advancements and applications, edge computing is already seeing widespread interest from industries ranging from manufacturing to healthcare and retail. Leveraging the growing power and ubiquity of CPUs and neural processing units, edge AI can process growing haystacks of data right where they’re being created, finding their needles quickly for local or remote processing.
Edge AI is an enabler for early networked autonomous cars — instantly recognizing and sharing details on accidents, weather conditions, and traffic from vehicle sensors and smart infrastructures in real-time. Similarly, edge AI has empowered wearables to actively monitor seniors for chronic health conditions, alerting remote caregivers within seconds of detecting abnormalities in their biometric data.
It’s clear that edge AI has the ability to open up a whole new world of insights and opportunities across multiple industries, but connecting the distributed data processors to usefully aggregate their discoveries is a higher-level task. That’s where a system of systems (SoS) comes in.
If you’re unfamiliar with the concept of SoS, you’re not alone: This relatively new frontier in edge AI computing seeks to connect an enterprise’s multiple, purpose-dedicated systems using a single common language. Currently, IT networks, manufacturing machinery, transportation assets, physical security, HVAC, and lighting systems each have their own communications protocols that weren’t designed to speak or integrate with others. A SoS serves as a superstructure, using AI to coordinate and aggregate data processed at the edge from these different systems.
The opportunities for SoS
SoS will allow autonomous or semi-autonomous systems to control and respond to data flows. In the defense sector, for example, it will connect the data dots gathered from weather analysis, radars, and video surveillance to provide either the quickest path for a missile, or the best way to intercept it. Separately, a train technology provider that delivers transportation as a service need to unify the subsystems in a train and in a train station, expediting failure flagging and repairs to reduce costly service delays. In each case, a system of systems will inform or replace human decision-making, leading to faster, smarter, and more precise insights.
It’s no stretch to say that edge AI-powered systems of systems will change society as we know it. Like bees working together to build and maintain a hive, algorithms in a SoS will form a swarm. Cars that can communicate with each other will be collectively smarter and safer than any individual car. Inside one vehicle, a SoS will coordinate navigation and telematics while independently gathering live weather and traffic data from roads. Then a multi-vehicle, infrastructure-level SoS will harvest that data across a fleet of connected vehicles, enabling dynamic map rerouting, automated emergency braking, and instant requests for assistance. We can also imagine a world where every mobile device can leverage the power of others to help provide the right decision at any time. This will be a revolution that will improve the lifestyle and even health recommendations for every single citizen.
Factory automation is another emergent use case. SoS will empower disparate machines on a factory floor to work with one another to create a finished product. One system will coordinate the robot molding a car door with window- and handle-creating robots, determining in real-time whether to ramp up door production, maintain the pace, or slow down. Manufacturers will be able to minimize part surpluses, reduce build times, and improve production outcomes.
Despite the myriad opportunities offered by SoS, enterprises still face challenges in turning this concept into reality. One of the biggest challenges is standardizing information flow, where the lack of standard data formats and communication protocols can be a fundamental problem.
For example, in today’s world a lot of individual vendors use their own data formats and protocols, requiring customers to use their proprietary software to understand data they are getting from devices and to depend on custom protocols for operational execution. This makes data integration expensive and reduces the ability of companies, factories, cities, etc., to build optimal machine learning models that can solve holistic problems.
Regardless of how robust vendors’ systems are, if they cannot communicate with one another in real-time, a city’s “smart” lighting system will fail. This is why standardizing data flow is so fundamental to realizing the real value of SoS, and to that end, hardware and infrastructure companies such as Siemens, Rockwell, and Honeywell are currently building system-integrating platforms.
Another problem that needs to be addressed is ensuring trust and privacy at the edge, as governance issues around data integration, privacy, and ownership are still not solved. Many “unknowns” have led to increased fears regarding privacy, some unfounded and others entirely reasonable.
Working with data requires establishing system-wide trust and convincing individuals that their data is being used and shared safely, particularly in industries such as healthcare, where data privacy and security are always priorities. On the other hand, powerful outcomes become possible when individual data is aggregated and analyzed at a higher level — exactly where SoS shines. How do we satisfy everyone’s concerns?
Denmark offers a glimpse into how SoS will impact the healthcare sector. In the modern era of precision medicine, one person’s genetic sequence consistes of 6 billion characters, and medical imaging now uses ultra-high-resolution 3D imagery. As a result, each patient’s record requires a large amount of storage, measured in terabytes, with extremely complex individual data sources.
Faced with electronic health data and privacy concerns, Danish health care providers are beginning to tie data directly to individual patients to provide security and quality of service while data are secured in large centralized cloud repositories. When patients move from one doctor to the next, their data goes with them, becoming temporarily available at the medical network’s edge via phones or computers, where it can be compared against the provider’s existing materials. This process requires rapid edge processing of large quantities of data spread across multiple different formats. Armed with patient consent, an individual doctor could share locally processed genetic condition findings with a medical cloud, either identifying potentially useful medicines, or contributing to a broader AI-scanned research database.
The future of data won’t be about aggregating all information in one place; it will instead use both local and big data to improve insights and inform subsequent actions. That said, these insights will only be as broadly based as the data fabric connecting their underlying systems allows them to be.
Making SoS work better for us
As companies focused on intelligent data processing continue to push the boundaries of what is possible with AI, we need to strike a balance at the edge so that SoS is possible. The big shift and smart decisions will come when people start seeing value from integrated systems data rather than siloed AI functions. Standardizing data and protocols is a central piece of the puzzle for some industries, while developing privacy and trust will be essential for others; in any case, attempting to innovate without meaningful distributed data management will prevent enterprises from reaping the net benefits of what SoS makes possible. The management of flow and integration of data across a distributed set of systems and sources will build the foundation for the “swarm algorithm” mentality that promises to make SoS so powerful.
SoS is still in early days, but efforts are underway to create common standards and platforms to harmonize data. In addition to the infrastructure platforms under development by Siemens, Rockwell, and Honeywell, public cloud providers are moving into the SoS space, focusing on consumer angles, public domains, and protocols, though they’re beginning to realize that the industrial and enterprise sectors are more complicated than they thought.
As standards emerge, understanding the opportunities to leverage SoS in your business is a good first step. The next step is to acknowledge and work to solve the data and communications barriers keeping you from drawing out its full potential. Once you’re informed, you can start moving toward connecting your data silos and get ready to open the door to the larger, transformative applications of edge AI.