Evaluation of Gradient Descent Algorithm for Attitude Estimation

Abstract

Estimation of object’s attitude plays significant role in many fields such as consumer electronics, robotics, and satellite missions. Here, an iterative method for attitude estimation based on the gradient descent algorithm is implemented and evaluated. Although the number of iterations and processing load of iterative algorithms are hardly predictable, which may represent a challenge in real-time applications, the presented approach provides flexibility in adjusting the complexity of the algorithm to a targeted embedded system. The relation simulation results show the between the maximum number of iterations, attitude estimation error and convergence rate, that can be set through several user-defined parameters according to the application requirements and computational resources.