Real-time on-board artificial-neural-network-based processing of sensory data

Wildlife studies are often marred with problems of accessibility and observability constraining data collection. However, with the advent of machine learning and, more recently still, edge computing, there is increasing potential for a step-change in elucidating the ecology of the globe’s wildlife.

Most ecological research follows a conventional data collection and processing paradigm, where raw data is first collected (for example by using bio-logging devices or camera traps) and then processed post-hoc (using a personal computer or server). Edge computing is a distributed computing paradigm that brings computation and data storage closer to the data sources. The main features of this computing architecture are reductions in latency and bandwidth use for data transmission.

The project aims to apply state-of-the-art machine learning algorithms (deep learning) on board sensory platforms to achieve increased, richer data collection at the source, better automation in ecological monitoring, and faster responses to disrupting events, such as weather conditions or disease transmission.

Progress 2023

A literature review entitled “Edge computing in wildlife behaviour and ecology” is under review in a scientific journal. In collaboration with the Wildlife Ecology and Conservation Group, the researchers of this project are testing the performance of a deep-learning framework on behaviour classification of multiple animal species using accelerometer data. This deep-learning framework can be configured on-board of animal tracking devices for automatic and long-term behavioural monitoring.

February 2024

In February 2024, the literature review entitled “Edge computing in wildlife behavior and ecology” was published in the journal Trends in Ecology and Evolution. In it, we discuss how modern sensor technologies increasingly enrich studies in wildlife behavior and ecology. However, constraints on weight, connectivity, energy and memory availability limit their implementation. With the advent of edge computing, there is increasing potential to mitigate these constraints, and drive major advancements in wildlife studies. 
The deep learning framework research will be finished and literature drafted in June 2024.