Overview

Helvar’s smart lighting infrastructure can best be described as an IoT framework where apart from enabling the core lighting functionality, the network is also responsible for sending data generated by luminaires, sensors, and other devices. An example of the evolution in technology is the development of a reliable high-speed data network which forms the basis of the analytics platform. Naturally, the platform also presents new challenges to be tackled as it scales up. This year we have defined three subcategories inside our challenge that present real-world problems related to smart buildings and intelligent lighting.  

First category

The first category in our challenge asks the question: how fast can we set up the network? Typical lighting installations in large buildings can contain hundreds and thousands of devices. The network begins to transmit data only when it has been installed and configured. However, it is important to know exactly where the devices are in the building in order to extract any meaningful insights. The current process is to physically go through every space in the building, identify the devices and locate them on a floorplan using a mobile application. This quickly becomes a time-intensive task as the size of the network increases. 

We would like to see if the participants can design automation algorithms that speed this entire process up. Instead of manually identifying all the devices, is it then possible to detect a few devices throughout the building and utilise sensor data to locate the rest? The basic assumption to be made here is that co-located sensors theoretically send similar measurements and therefore the data should enable us to approximately locate a cluster of devices. Ideally, we should be able to reduce the number of devices which need to be identified while achieving the highest accuracy in determining the locations of the rest of the devices on the floor. The dataset will include fully anonymised floorplans with all device locations and time-series of their events. We will also provide an additional test dataset where the participants can showcase their approaches. 

 

Second category

The second category is all about creating applications and services for smart buildings. Let’s assume that all devices have been identified with their locations on a floor plan. We would like to see if it is possible to detect the busiest and quietest paths in the building. This is a very useful example of how we can use sensor data to understand building dynamics. In large buildings for example, the flow of the occupants could change according to the time of the day, which allows our clients to gain a deeper understanding of how their premises are being used. We will provide a similar dataset to the previous category, which will contain the location and events from all the sensors on a floorplan.  

Third category 

For the last category in our challenge, we have recorded data at Helvar HQ’s garage. On a busy day, we observe normal vehicular traffic flow in and out of the garage as well as employees riding their bicycles to work. People also tend to walk around in the garage when entering or exiting from their vehicles, to go to the nearest elevator or stairwell. The garage at Helvar HQ already contains a network of ActiveAhead devices in which each luminaire is equipped with a motion sensor. We decided to augment the set up by placing multiple audio sensors around the garage to capture a second stream of information as well as motion data. The challenge is to identify what is actually happening in the garage. We call it the ‘Mystery Garage’.  

Remember, even though we would like to see strong accuracy scores on datasets, it is more about the quality and innovativeness of the approach and how you reason about your algorithms and present the results. We are excited about what the participants will come up with. Good luck!