Sistem Monitoring Electric-Powered Wheelchair (Epw) Berbasis Sensor Accelerometer Terintegrasi Kalman Filter Dan Auto- Encoder Machine Learning

Abstrak

Abstract
In their daily lives, people with disabilities require a wheelchair as a mobility aid. Over time and with technological
advancements, the improvement of wheelchairs for people with disabilities has not only focused on quantity but also on
enhancing functionality. The development of control systems includes the use of direct joystick-based controls and the
modification of manual wheelchair propulsion systems to electric. The use of Electric Powered Wheelchairs (EPW) needs to
be supervised by a third party to ensure the safety of people with disabilities. However, in certain conditions, individuals with
disabilities require personal space, and continuous direct supervision by a third party may not be feasible Monitoring systems
can assist in supervising people with disabilities, but the accuracy of signals from the accelerometer sensor, the foundation of
the monitoring system, needs careful consideration. This can be addressed by integrating Kalman Filter and Autoencoder
algorithms. To evaluate these algorithms in this study, EPW testing was conducted on track patterns 1 and 2. The obtained
accuracy values were (1) without integrating Kalman Filter and Autoencoder Machine Learning, 16.84%, (2) with Kalman
Filter integration, 93.06%, (3) with Autoencoder Machine Learning integration, 95.59%, and (4) with Kalman Filter and
Autoencoder Machine Learning integration, 98%. This indicates that the algorithms used to reduce noise from the output data
of the accelerometer sensor have good capabilities in reducing noise from the accelerometer sensor's output data.
Keywords: electric powered wheelchair, monitoring system, sensor MPU6050, kalman filter, autoencoder