RIOT-ES

Developing technologies for Resource-Efficient IoT-Edge Systems

Riot-ES project gathers four partner organisations to research and develop new methods, technologies and systems for maximising energy efficiency and increase performance in IoT systems, with a focus on IoT devices and processing at the edge of wireless networks.

Energy efficiency and performance are often at odds with each other, and the project will investigate fundamental trade-offs between the partners to understand how to optimise energy efficiency given performance constraints and conversely, how to optimise performance given energy constraints. The findings will be used in two different use cases - Smart Cities and Smart Home - of specific commercial interest to the partner organisations, where they will generate commercial advantage in their respective market segments.

We will explore a realistic urban use case on Smart Parking, our low-cost, seamless and straightforward-to-install parking system. It incorporates vehicle-detection electromagnetic sensors, an intelligent software platform to collect and process all the data, along with web and mobile applications (respectively for parking managers and drivers). On one hand, it offers drivers access to real-time parking occupancy information, reducing the time each driver spends looking for a parking space. On the other, it supports parking operators on improving their operations by offering key performance indicators in real-time and reports on behavioural patterns detected by the platform.

This use case will implement and demonstrate the addition of a video camera to monitor in real-time the available slots in a parking facility. The media acquired from the camera will be processed using image recognition software to allow the perception in real-time of the status of each parking spot.

Different AI algorithms with different computational costs are aimed to be used in the context of this use case, as to achieve the lowest possible energy consumption.

In the envisioned deployment, cameras will be connected to one or more edge nodes and stream in real-time its acquired media data. The edge nodes will pre-process the data with lightweight image recognition algorithms. This way, by avoiding sending all media data to a cloud central node responsible to fully process all the streams acquired from a smart-parking deployment, there will be savings on network consumption resources as well as on the computational power needed to fully process all these streams in real-time.