Generalized Nonlinear Complementary Attitude Filter

A new attitude estimation technique that is a generalization of Mahony's nonlinear complementary filter is presented. The two primary attitude representations used are the direction cosine matrices for a global representation and the Euler vector for small changes in attitude near the identity. No assumptions are made as to the dynamics of the rotating object. The nonlinear complementary filters combine low-frequency attitude information from a set of vector measurements with high-frequency attitude information from rate gyros. This generalized attitude filter gives the filter designer an enormous space to tune and optimize the filter while ensuring stability. The generalized attitude filter does not include as a special case the full multiplicative extended Kalman filter (MEKF), which may not necessarily have a positive-definite integral gain.