The fourth presentation in the A2IR2 webinar series was successfully delivered by Mr. M. I. Eghrudje on Wednesday, 17th May, 2023.
Titled “Determination Of Finger Pose for A Wearable Hand Orthosis Using Kalman Filtering”, the talk was aimed at showing how Kalman filtering as a recursive digital tool can be integrated into a robotic hand orthosis to provide far more precise and accurate estimation of finger orientations and positions based on readings from the IMU(Inertial Measurement Unit).
Accelerometry time series data from Inertial Measurement Unit (IMU) are plagued by high frequency noise. Such high frequency noise can sometimes be reduced using low pass or moving average filters. However, for use-cases such as the one that is the subject of this study, such traditional filters cannot capture dynamic changes in the underlying pattern of the data. There is thus a need for a filter capable of automatically capturing such dynamics, which motivated the use of Kalman filters in this study.
To determine the combined pose of the fingers for a robotic hand orthosis, MPU6050 sensors were utilised as IMUs, with the goal of providing information about the fingers’ motions and orientations in 3D space. Each IMU contained a three-axis accelerometer and a three-axis gyroscope, which allowed it to measure both linear acceleration and angular velocity along three axes. A Kalman filter was used to implement a low pass filter to remove noise from the accelerometer data. It optimised the handling of noise in the system by estimating the state of the system and the noise simultaneously. This allowed the filter to distinguish between the signal and the noise, resulting in better noise reduction. Kalman filter served as an optimal estimator. This means that it used a probabilistic model of the system dynamics and measurement noise to estimate the state of the system with the minimum mean squared error (MMSE). The MMSE is a measure of the difference between the estimated state and the true state of the system, averaged over multiple estimation cycles. The Kalman Filter minimized the MMSE by continuously updating the state estimate based on new measurements and predictions of the system dynamics. With this Kalman filter is able to provide the most accurate estimate of the system state possible.
The talk therefore provided an explanation of the Kalman filter algorithm, the mechanism used in the construction of a suitable hand orthosis as well as an overview of the entire control system. Comments from members of the audience showed that the talk was very well received. The recording of the seminar can be obtained here: https://www.youtube.com/watch?v=9x_Vz1mTo0s&t=1360s.