- Main
Sensor-Based Attitude Estimation of the Human Arm
- Kannanda, Vikas Chinnappa
- Advisor(s): de Oliviera, Mauricio
Abstract
Wearable technology has grown rapidly in popularity in recent years. The advances in accessible micro-controllers, affordable sensors, and high-level interface design suites has made this possible. At the same time, interfaces for gestural control have become more attainable. A concurrent trend in DIY electronic instruments and controllers has evolved for similar reasons. In combining these phenomena, and in searching for greater control of live graphical interfaces, several people have developed wearable controllers in the form of exoskeletons and gloves. In general, the impetus behind wearable controllers, and more specifically optical tracking, is to impart the ability to control graphical interfaces through gestures.
Optimal state estimation theory is used in this thesis to build a linear model of the arm which by default has high degree of non-linearity. By using Kalman filtering, we can estimate the complex dynamics of a non-linear system with the simplicity of a linear system. First a model which contains the dynamics of the arm is introduced as the plant. After some definitions and derivations of filtering are established, the position is then estimated using an optimal filter.The use of a model that introduces the torques applied as a state variable instead of an input greatly improves the estimation.
The thesis is concluded with a design of a lightweight hardware architecture that can be used to implement the estimation. The estimation is found to deliver fast tracking of the highly non-linear system but breaks down at certain points due to the nature of Euler angles producing Gimbal lock. To overcome this the concept of quaternions algebra is studied and basic position calculation is done using quaternions.
Main Content
Enter the password to open this PDF file:
-
-
-
-
-
-
-
-
-
-
-
-
-
-