AIoT on smart cameras has the potential to create innovative solutions in many domains. These cameras' capabilities can be paired with other emerging or existing tools and technologies to help create innovative solutions. Therefore, IoT-based analytics will be one of the focus topics in our February 2023 research report, Supply Chain Data And Analytics State of The Market. In this article, we share an example of how an existing real-world AIoT solution can be paired with simulation modeling to help create an "always-on" advanced analytics capability. I recently came across a use case by a digital video solutions provider has its video solutions deployed across thousands of locations in India. Edge devices on the solution leverage an Intel compute stick that provides a powerful platform with an Intel Core M Processor and integrated GPU for processing video. The stick is powered by Intel Movidius Myriad X VPU, which is already being leveraged across millions of smart security cameras and edge devices across the globe. The critical aspect of this solution is that through the combination of hardware and software solutions, the solution can provide a robust set of analytics, prebuilt within the solution, like:

• Walk-in counter
• Audience metrics
• Visitor path mapper
• Recognizer
• Queue counter
• Dwell counter

The smart camera captures and relays metrics and measures like these to the cloud. The above list is not exhaustive but is indicative. While these numbers are undoubtedly essential and insightful, the value of these numbers can only be extracted by leveraging them for various types of analytics. As an example, one area where value can be created is by combining these streaming metrics with other forms of real-time analytics. Let us use the example of simulation modeling in the context of Queuing theory. The conventional queuing theory formulas used in classic operations research are constrained by many assumptions. However, simulation modeling allows us to erase those constraints. I assume you are familiar with queuing theory or how queues or flows in public places are studied and optimized. In that case, you can already see some of the metrics captured by the camera are useful. With streaming, near real-time data, we no longer have to make assumptions on the probability distribution of inbound traffic, the distribution of wait times, processing times etc. There will be aggregations in the model, but with the continuous streaming data feed, you can simulate as close to reality as possible and plan accordingly. The improved accuracy in traffic management in retail can have significant cost benefits, ranging from better product assortment and placement to reduced labor costs. Store-level profitability can be significantly improved. Another area is warehouse operations. Generally, warehouses have measured KPIs like units per man hour (UPMH) leveraging the RFID or barcoding scan transactions from systems like SAP. While these scans and subsequent transactions in the system are also near real-time, building a real-time analytics system that is constantly being fed these transactions and then aggregating and calculating metrics, though feasible, is not as practical and accurate as the scenario where smart cameras, optimally placed across the warehouse, not only capture fundamental KPIs like UPMH, but can also help regulate labor scheduling, among many other aspects of warehouse operations planning. This is an opportunity for platform solution providers like SAP BTP to help build this aspect into their industry solutions. For some industries, this can be done relatively easily and will be a major differentiator from a product offering perspective. Opportunities are many and we are constrained only by our imagination and willingness to erase the constraints that in reality, no longer exist.