Jordan Cavendish

Image

Snapshot of power in an array of seismic sensors as a giraffe walks through them.

My research investigates how distributed sensor networks can be used to monitor wildlife through localisation, tracking, . In particular, I am interested in seismic sensing, an under-explored modality that can detect the vibrations generated by animal movement, enabling tracking without relying on visual line-of-sight sensors or animal vocalisations. Current wildlife monitoring systems often focus on isolated tasks such as detection, classification, or localisation, with each stage treated independently. My work aims to develop machine learning methods that combine these stages into an end-to-end tracking framework, directly from sensor data, allowing information from detections, identities, and trajectories to inform one another. To support this, I have developed a seismic footstep simulator that provides synthetic data with complete ground truth for evaluating localisation and tracking algorithms.

Image

In the future, I intend to investigate how data from multiple sensing modalities can be fused to improve localisation accuracy and robustness, as well as the ways that animal behaviour can inform and be incorporated into tracking models. Ultimately, the goal is to enable fine-scale, non-invasive monitoring of animal movement and behaviour in natural environments. Such methods could provide new insights into ecology and conservation while reducing the need for invasive tagging or direct human observation.