Hi all,
I have developed a new deterministic attitude estimation algorithm which when coupled with Mahony's nonlinear complimentary filter gives really good results. The algorithm uses 3-axis accelerometer and 3-axis gyroscope to do the computation but is generalizable to a "vector correspondence" and 3-axis gyroscope. A vector correspondence is defined as a set of two vectors which are the resolution of a single free vector resolved in two different coordinate frames.
A Matlab implementation of the algorithm has been uploaded to Bitbucket and a video demonstrating the attitude estimation's robustness can be found on Youtube. In the experimental validation, we use the fact that a spherical robot's relative position is completely determined by the attitude. Hence, if the estimator estimates the attitude correctly, we should be able to estimate the position of the spherical robot.
I am also thinking of comparing the performance of the algorithm with the DCM algorithm suggested by William Premerlani and Paul Bizard. My journal paper, published recently in IEEE Transactions on Mechatronics, compares the algorithm with the TRIAD and TRIAD+Mahony's complimentary filter. DAESR algorithm out-performs these algorithms in both accuracy and robustness. You can find a preprint of this paper at my webpage.
The reason for posting this: I do not have access to flying drones and would like to see the performance of this algorithm under flying conditions. I am envisioning a setup where the attitude of the drone is logged along with the sensor data (accelerometer and gyroscope) and an offline comparison of the existing attitude estimation with DAESR estimation.
If you are interested to help me out, please contact me at aby[dot]vinod[at]gmail[dot]com.
Cheers,
Abraham
Webpage: www.unm.edu/~abyvinod
Replies