Affirmo: Could AI make a reality of IoT dreams?

Lip Sing Tay is both a busy and accomplished individual, steadily turning the promise of the Internet of Things into a reality from the base of his company's operations in Singapore. The advent and popularisation of AI, Lip Sing related in an email interview with TechDay, holds considerable potential in accelerating the performance and utility of granular tracking and insight systems across supply chains.

The former Head of IBM's Data Management, Analytics and AI for South Asia says Affirmo is the result of his direct experience of these technologies being implemented in the enterprise data centre space. But there was a problem: IoT stubbornly remained more promise than reality. "I observed that despite forecasts of the growing IoT market, there are very few real-life successes," he says.

As he studied the market further, Lip Sing appreciated the difficulty faced by any potential IoT user to realistically achieve a complete working tracking solution. The problem? Complexity. "It requires technology [and expertise] from multiple domains - big data, data management, analytics, wireless, radio frequencies, IoT, software development, and more," Sing Tay explains. "End organisations [therefore] struggle to implement good tracking solutions and as a result, many still do not have real-time location intelligence capability."

That made his mission clear: deliver robust, scalable and complete solutions for tracking people, assets, and vehicles – including in some of the world's toughest environments. Doing that might seem like a task for cellular networks, but as popular and even ubiquitous as these are, that's not where Affirmo and Lip Sing looked.

He explains that among the pre-eminent challenges particularly for tracking individuals (generally for security purposes) is physical sensor mass, along with mobile power: batteries are heavy and require charging. "We use highly energy efficient and long range LoRaWAN sensors and protocol, which uses lightweight trackers with exceptional battery duration, and which ranges across kilometres" says Tay Sing, adding that other (somewhat exotic) forms of wireless networks are also used, including NB-IoT BLE RSSI (Bluetooth Low Energy Received Signal Strength Indicator), Ultra-Wideband (UWB) or BLE AoA (Bluetooth Low Energy Angle of Arrival), and where feasible, standard cellular networks.

Noting that tracking multiple thousands of moving objects in a production environment comes with its own challenges, Lip Sing says billions of over-the-air signals are processed sequentially to provide precise spatial and temporal positioning. Affirmo's location engine is capable of tracking more than 100,000 moving trackers at one time. "As such, our enterprise customers can rely on us to have correct location information for safety, security and compliance. Having correct data round-the-clock gives them peace of mind."

This range of technologies, says Lip Sing, overcomes the limitations of GPS (which doesn't work indoors), as well as the aforementioned battery and sensor size issues. "We deploy a range of technologies, finding the best to solve our customer's situation and problem."

Readers will note that while IoT is conceptually younger than AI (with its roots at Carnegie Mellon and an ARPANET connected Coca Cola machine in 1982), it has been 'in market' at some sort of scale for considerably longer. However, Lip Sing's opening point stands: IoT has perhaps not yet achieved the sort of universal connectedness and granular supply chain visibility initially envisaged.

Which begs the big question: Could the combination of AI with IoT finally unleash fuller potential? Lip Sing hints that this may be the case, though he's not revealing all his cards immediately. "Our real time location tracking solution provides a user interface, customisable workflow, advanced data analytics and even AI. Simply put, Affirmo provides a complete end-to-end stack, covering location data, software and vertical solutions."

He adds that AI excels at finding patterns in large amounts of seemingly unrelated data. "AI makes it easier to discover hidden patterns, anomalies, predictive trends, and inefficiencies in IoT data. We can see AI already being deployed in IoT domains such as predictive maintenance, sustainability and security, and we will continue to see more advances in the coming years."

He points to the example of a real time asset tracking in manufacturing. "In addition to location of work-in-process [WIP] items, our analytics provide the exact duration a WIP item spends at each station, and in the whole manufacturing process. This can be used for further analysis to further optimise the manufacturing process. Our advanced analytics can also further suggest which completed WIP item should be selected depending on the manufacturing customer's inventory policy such as First-In-First-Out (FIFO)."

By using AI on our location data, he says manufacturing companies can benefit from predictive scheduling, planning and optimising inventory, manpower and other resources.

There is, as one might imagine, more to come as AI and IoT inevitably draw closer, and Lip Sing confirms that Affirmo is working on additional vertically integrated tracking solutions helping customers optimise their use of tracking data. "There are many opportunities for AI to be applied into our location intelligence analytics and data, for example, in anomaly detection, flagging impossible or suspicious movement patterns, such as a person 'teleporting', an asset moving through a blocked route, or a tag staying still too long in a high-risk area," he says.

The secret, of course, is that customers themselves are likely to see the best use cases for ever-vigilant AI to go to work on their logistics data and expose new value from it.

There could, in other words, be a new wave of innovation emerging from the IoT space thanks to AI – and Affirmo is positioning itself to be at the forefront of enabling its customers to do just that.