Quaternionic Attitude Estimation for Robotic and Human Motion Tracking Using Sequential Monte Carlo Methods With von Mises-Fisher and Nonuniform Densities Simulations

Abstract

In recent years, wireless positioning and tracking devices based on semiconductor micro electro-mechanical system (MEMS) sensors have successfully integrated into the consumer electronics market. Information from the sensors is processed by an attitude estimation program. Many of these algorithms were developed primarily for aeronautical applications. The parameters affecting the accuracy and stability of the system vary with the intended application. The performance of these algorithms occasionally destabilize during human motion tracking activities, which does not satisfy the reliability and high accuracy demand in biomedical application. A previous study accessed the feasibility of using semiconductor based inertial measurement units (IMUs) for human motion tracking. IMU hardware has been redesigned and an attitude estimation algorithm using sequential Monte Carlo (SMC) methods, or particle filter, for quaternions was developed. The method presented in this paper uses von Mises-Fisher and a nonuniform simulation to provide density estimation of the rotation group SO(3). Synthetic signal simulation, robotics applications, and human applications have been investigated.

Survey of Motion Tracking Methods Based on Inertial Sensors: A Focus on Upper Limb Human Motion

“…The major drawback of such a solution (and of all the three parameters attitude representations) is the existence of singularities that may occur in certain configurations, as in the case of the gimbal lock. This is why many methods use quaternions for the estimation (e.g., [38,39,40,41,42]). Quaternions allow for a computationally efficient and singularity-free attitude representation and are often used for the attitude representation of a single body.…”

“…The method proposed in [70] uses unscented transformations, but implements constraints through probabilistic graphical models. Particle filters have been used by [39,42] to further improve and generalize the representation of PDFs.…”

“…One example is the work of Zhu et al [32], in which the authors constrain the hand to follow a straight line and then check the reconstructed trajectory to be straight. Other mechanical platforms and robots were used in [42,57,75,76,81] as ground truth data.…”

“…This second part can be adapted depending on the availability of measurements; e.g., acceleration due to gravity in case of an IMU and, in addition, local magnetic field in the case of an mIMU. Finally, the method of To and Mahfouz [42] tries to improve the quaternion attitude estimation by using von Mises-Fisher and Bingham densities in a PF that provides the attitude quaternion based on the IMU signals.…”

Generalized Linear Quaternion Complementary Filter for Attitude Estimation From Multisensor Observations: An Optimization Approach

“…T HE development of consumer electronics brings a worldwide mania in cellphones, smart wearables, interactive devices and etc. [1], [2], [3]. Such electronic products indeed improve the quality of our living.…”