This thesis focuses on examining flapping flight models, and the possibility of their introduction in behavioural algorithms, all while still being able to run in real time. For this purpose, we present a detailed comparison between flapping flight models that are with today's hardware capabilities in mind best suited for the development of a comprehensive physical system of flying birds. We present their implementation specifics as well as the advantages and limitations.
More specifically, we first present an overview of bird flight mechanics, the complexity of 3D bird models suitable for the automatic generation of diverse kinds of birds. We continue by presenting a model for calculating aerodynamic forces with the use of simplified equations and suggest the integration of stable PD controllers. In the context of our problem, we then describe four different, but most appropriate approaches of physically based simulations of flapping flight: a) a visually convincing model, b) a simplified model, c) a model with consideration of airflows and d) a data-driven model. In the discussion part of this thesis, we introduce a type of a learning flapping flight model capable of steering in a 3D space, which works based on a pre-defined path and a heuristic representation of the desired orientation. With the intention of making the controller appropriate for behavioural algorithms, we introduce a procedure for combining strokes, without unnecessarily large data storage. Following the need of a robust flapping controller, we propose a wingbeat database that is able to manoeuvre under different angles of attack. We conclude by presenting possible improvements and ideas for further work.