Talk on "Attitude estimation in autonomous six Degree of Freedom"
Speaker: Dr. Yujendra Mitikiri, Ph.D. in Mechanical Engineering from University of Florida, Gainesville.
Title: Attitude estimation in autonomous six Degree of Freedom
Abstarct:
In this seminar, we shall consider the problem of attitude (or orientation) estimation in small, resource- constrained, autonomous vehicles. In particular, we are interested in the development of a solution that
could be carried on board a small unmanned aerial vehicle (UAV), or autonomous underwater vehicle
(AUV). The motion of aerial and underwater vehicles encompasses the complete six dimensional transla-tional and rotational degrees of freedom, unlike surface vehicles which primarily experience two translational
degrees of freedom, and one rotational degree of freedom. This makes the problem of attitude estimation and control more challenging in aerial and underwater vehicles.
Existing solutions for attitude estimation fall under two flavours: a purely geometric problem involving
vector measurements, and a geometro-kinematic problem involving angular velocity and vector measure-ments. The former type have analytic solutions for a quadratic output error optimality criterion. The
latter type provide asymptotically convergent solutions.
We shall show that both kinds of attitude estimation problems may be treated under a single unifying
geometric framework. We present theoretical solutions in closed form using this new approach for both problems. The proposed approach also supports a wider variety of optimality criteria for the geometric
problem, while yielding faster and more accurate solutions for the geometro-kinematic problem.
Preliminary experimental work has yielded promising results for the estimator’s performance in com-parison with reported work, for pure rotational motion under laboratory conditions. We are now working
on implementing the estimator on a UAV and verifying its performance in real flight, where the estimator is subjected to both translational and rotational motion.
Practical implementation of a theoretical solution generates its own set of problems, which we enumerate and propose methods to attack and solve. The inertial sensors are noisy and possess bias, which need to be
filtered out and compensated for respectively. This is relatively a straightforward task. The measurement of vectors is also error-prone under practical conditions with translational accelerations and external magnetic
fields, and compensating for these errors is typically more challenging, if not actually intractable. We propose the incorporation of a low-level visual sense to provide an accurate correction of the residual errors
in the attitude estimated using the non-ideal inertial sensors. We shall finally validate the implementation of the attitude estimator on a real UAV.
Speaker Bio:
Yujendra Mitikiri received his B.Tech. degree in Computer Science and En-gineering from the Indian Institute of Technology Madras, Chennai, in 2001. He worked as an analog circuits engineer at Texas Instruments from 2001 to 2015.
While at Texas Instruments, he was granted four patents at the US PTO for novel circuit techniques that help in the design of faster, more accurate, and more efficient data converters. He was elected Senior Member Technical Staff
at Texas Instruments, before he resigned in 2015 to pivot to a competely new engineering field. He received his M.S. degree in Aerospace Engineering from the Department of Mechanical and Aerospace Engineering at the University of
Florida, Gainesville, in 2017, and most recently his Ph.D. in Mechanical Engineering from the same university in 2020. He has published three journal articles and presented at three
conferences on topics in the field of control and estimation in autonomous six degree of freedom robots while pursuing his doctorate degree. His research interests include autonomous six degree of freedom robotics,
control and estimation, nonlinear and geometric methods, sensors and instrumentation, and analog circuits and systems.
Event Date: 02nd July, 2020(Thrusday)
Event Time: 03:00 PM