A University College Dublin (UCD) researcher, Professor John Murphy, UCD School of Computer Science, is advancing state-of-the-art, cognitive-edge computing to enable smart campus applications, through an IBM faculty award.
Edge computing applications, such as UCD’s 'smart' campus applications, use the processing power of edge devices to filter, pre-process, aggregate or score Internet of Things (IoT) data while harnessing the power and flexibility of cloud services to run complex analytics on the data, and in a feedback loop, support decisions and actions.
Value of extending computing and cognitive service capabilities
Through the IBM faculty award, Prof Murphy is showcasing the value of extending computing and cognitive service capabilities to the edge while demonstrating tangible improvements in key performance indicators.
Prof Murphy’s team at UCD’s Performance Engineering Laboratory (PEL) is developing two 'smart' campus applications using an IoT platform pilot within the UCD School of Computer Science building in collaboration with UCD’s real estate services.
This platform is composed of a dozen edge devices such as system-on-chip Raspberry Pis, which are wirelessly connected to an edge gateway, connected to the IBM Bluemix cloud service. Each Raspberry Pi is equipped with multiple intelligent sensor devices including temperature, humidity, atmospheric pressure, accelerometer and gyroscope.
In the first use-case the PEL team is using edge devices to monitor the temperature, relative humidity and barometric pressure in various locations within the school building.
Collected data used to generate heat maps and humidity maps
In the IBM cloud, the collected data is used to generate heat maps and humidity maps which, in turn, are being used to detect heat loss in specific building zones enabling a more cost-effective facility management service.
Prof John Murphy said: “The information from this use-case is valuable for UCD campus real estate managers, allowing a more efficient way to monitor the building’s heating system. While this has been tried before, UCD has been entirely rethinking the way this is effectively offered through the use of edge computing.
"In addition the processing capability of sensor streams at the edge brings enhanced fidelity of sensor-induced insights, at lower cost, when compared to a centralised cloud-only approach.”
The second use-case application is focused on detecting human activity in approved rooms within the building.
The UCD PEL team is working on building models that ingest unstructured and structured sensor data feeds on the edge to detect aggregate human presence within a space.
Increasing fidelity in sensor stream processing
Prof Murphy said: “This work is extending our understanding of aggregate human activity more broadly. The novelty is in increasing the fidelity in sensor stream processing within cost boundaries, looking to bring it to new levels of efficiency, when compared to a cloud-only approach.”
Through this collaborative research programme, the UCD PEL team is leveraging IBM cloud services and the IBM Watson IoT Platform to facilitate quick and easy application development, deployment and operation on the edge.
Prof Murphy said: “Overall, the collaboration with IBM Research Ireland is pushing cognitive service intelligence closer to the data sources of the IoT. By doing so, basic data analysis is performed closer to the edge of the Internet of Things. More advanced and broader-visibility data analysis can then be performed in the cloud.
"This approach not only relieves the stress on the core network – as only selected and pre-processed information data is conveyed to the cloud – but it also enables deploying latency-sensitive responsive applications and services at the edge of the network, right where they are needed.”
Prof Murphy and his team at
UCD’s Performance Engineering Laboratory have collaborated with Yiannis Gkoufas, IBM research software engineer, as part of the IBM faculty award.