Engineering a machine learning system for use off-grid
Sensors are a great way to collect information in the field. Traditional sensors operate as data loggers whose results must be collected periodically; more recent devices have used wifi to return data automatically, but this requires infrastructure that’s typically unavailable when the sensors are really off-grid. We’re working on a sensor design that combines on-device machine learning and long-range/low-power radios to collect information about bird species abundance. This approach (sometimes called “TinyML”) increases the amount of computing done at the edge and reduces the centralised compute, network bandwidth, and infrastructure requirements.
Keywords
sensor networks, ecology, ai, machine learning, tinyml
Staff
Simon Dobson, Ali Johnston, Laura Aitken