Attitude determination from single camera vector observations

Abstract

Attitude determination is of major importance in Guidance and Control Systems of the Unmanned Aerial Vehicles (UAV's). Supplying wrong or not precise attitude very often turns to be catastrophic for the UAV's. Vision sensors are must nowadays. They provide reach source of information given as relative measurements between the vehicle navigation parameters (position, velocity and attitude) and the environment. This paper presents a framework for attitude determination from single camera vector observations. We assume known environment in a form of a map and true vehicle positions from which each observation has been taken. Two different methods for attitude determination are presented: an iterative numerical solution based on Gauss Newton's method and an exact method known as the Davenport q-method. Pros and cons of the both solutions are presented.