Analysing the Motion of Biological Swimmers

点击翻译按钮获取中文摘要

Consider a deformable object moving in some medium. In general, one cannot infer the internal deformation of the body based on surface shape alone: it requires detailed knowledge of the mechanical properties and forces operating at every point of the object. And yet, such inverse problems are of considerable practical and fundamental interest in soft matter physics, material science and engineering. We propose that the relatively simple and constrained geometry of the worm C. elegans, and our extensive knowledge of its anatomy and material properties makes the inverse problem tractable. A new and promising approach to the solution of inverse problems is provided by deep learning, which has already transformed performance on standard tasks from across artificial intelligence, particularly in the areas of language and vision, and increasingly now in robotics. The proposed research is to explore the feasibility of using deep learning to understand the internal and external forces acting on the body of C. elegans from video footage of the worm swimming in three dimensional complex fluids.