Modelling and Inference for the Movement of Interacting Individuals

这个项目将着眼于多个被跟踪的野生动物的连续时间运动模型，例如使用GPS，其中跟踪在时间上重叠，允许推断个体之间的相互作用。重点可能放在数量相对较少的个人身上。一个关键的工具将是最近开发的一种用于状态切换扩散模型的推理技术--表示运动和行为--其中状态之间切换的概率取决于位置。这种方法被称为“统一化”，其基础是将感兴趣的过程与众所周知的更简单的过程联系起来，该过程在空间上是一致的。该项目将通过研究更高维度空间中的扩散来扩展这一方法，以同时表示多个个体，并在具有更复杂结构的状态空间中表示更复杂的行为。还将开发加速这些模型拟合的算法。没有正式的外部合作伙伴，但与动植物卫生局的生态学家进行了非正式合作，包括获得关于野猪的GPS数据。

This project will look at continuous-time movement models for multiple tracked wild animals, e.g. using GPS, where tracking overlaps in time, allowing inference about interactions between individuals. The emphasis is likely to be on relatively small numbers of individuals. A key tool will be a recently developed inference technique for state-switching diffusion models - representing movement and behaviour - where the probabilities of switching between states depend on location. The method, known as "uniformization", is based on relating the process of interest to a well-understood simpler process which is uniform in space. The project will extend this methodology by looking at diffusion in a higher-dimensional space, to represent multiple individuals simultaneously, and at state-spaces with more complex structure, to represent more sophisticated behaviour. Algorithms to speed up the fitting of these models will also be developed. There are no formal external partners, but there is informal collaboration with ecologists at the Animal and Plant Health Agency, including access to GPS data on wild boar.