Inferring Rules of Motion in Animal Groups
Our project aims to discover the individual behavioural rules that determine the pattern of collective movement in animal groups. Inferring these rules involves statistical inference analysis on experimental data collected from individual animals as well as both large and small groups to determine the most probable form and strength of inter-individual interactions as well as the simulation of animal groups to determine if the proposed rules are able to reproduce the observed group behaviour. The project focuses on collective movement in both avian (pigeon) and aquatic (prawns, fish) groups and uses state of the art methods for animal tracking, such as GPS and computer image analysis. My specific role in this collaborative project is in the adaptation and application of inference methods developed by the machine learning and control systems community to construct the most plausible model for the systems under investigation by learning the parameters and the structure of the model directly from the data, the so-called 'inverse problem'.
The project is a collaboration between Richard Mann, CIM, Andrea Perna, Daniel Strömbom and David Sumpter at the Dept. of Mathematics, Uppsala University in addition to our colleagues in Berlin, Oxford and Sydney.