For wearable and connected devices, the pet market is an interesting and increasing one. Whether it is to assess your pet’s health, activity, or safety, various IoT technologies can be employed, from RFID sensors, to GPS trackers, to Bluetooth. IoT News reviewed one such product, PoochPlay, back in 2018.
It is the latter which PixelPlex, a development and consulting company focused on blockchain, artificial intelligence (AI) and IoT technologies among others, used in a project with an industry leader. The project is for a wider enterprise IoT mobile app and data gathering solution, but has an example use case for tracking pets’ activity and motion data.
Users are able to collect their pet’s motion data, with the mobile app allowing them to adjust collection sets and make custom fields, such as name, weight, height, breed, before data capture.
The IoT solution’s main purpose was ‘not simply to collect various motion data sets, but to extract specific types of information from all collected data’, PixelPlex notes. ‘It was important to enable this process before data upload for far more complex use such as motion pattern modelling, machine learning, and data science’, the company adds.
The solution has three steps, and utilises Bluetooth Low Energy (BLE) technology. The motion sensor firstly captures and sends data which comes primarily from an accelerometer and gyroscope. This is the data collection setup phase. The second phase concerns data set capturing; this can include any motion patterns, such as velocity, duration and frequency. Alongside data from the two main sensors, collection sets can also include GPS-driven data, pictures, videos, and sound files.
The third phase involves uploading data packets. When a collection pack is ready, the app checks if the recorded data quality is adequate before uploading to the cloud, or suggesting another capture needs to take place.
“This was quite an interesting project,” explains Alexei Dulub, founder and CEO at PixelPlex. “The idea was to attach sensors and integrate the data from those sensors, and use it to track paths and generate lots of data sets for further processing by the AI, creating machine learning data sets. In essence, it was a behavioural tracking mechanism.’
“The main challenge was to compress all this data, because there are limitations of how much you could transmit, and creating this mechanism to gather as much data as possible and to transmit it as quickly as possible,” adds Dulub.
BLE’s low power consumption and low battery usage are positive aspects of the technology, but it can cause limitations in terms of limiting the transfer ratio for the data. PixelPlex used ‘various transmission and compression algorithms’ to overcome this issue. “We found a couple of technical approaches in how to pack everything on the fly, send it over to the cloud and unpack it there, and apply all the machine learning algorithms for the data,” explains Dulub.