Localization of a Smart Robotic Walker using encoder-based odometry techniques, Inertial Measurement Unit and sensor fusion with Extended Information Filter
Robotic Smart Walker, Localization, Odometry, Extended Information Filter, IMU, Encoders, Kinect.
This work proposes the improvement of the prototype of a robotic smart walker aimed at assisting people with reduced mobility in physiotherapy rehabilitation procedures. The device consists of the adaptation of a conventional walker, to which were integrated gear motors, an Arduino Mega microcontroller, a BeagleBone Blue microcomputer, in addition to sensors such as incremental encoders on the wheels, a Kinect camera, and the BeagleBone integrated IMU. The focus of the work is to solve the problem of localization of the walker, considering the need to obtain an accurate estimate of the position and orientation of the device. The initial approach involves the use of odometry based on encoder readings and the IMU, with calibration to minimize accumulated errors. Data fusion will be performed using the Extended Information Filter (EIF), the dual of the Extended Kalman Filter, aiming to integrate sensor information efficiently. Recognizing the limitations of odometry, such as the accumulation of errors over time, the work will also investigate the inclusion of exteroceptive sensors to improve the robustness of the localization system. To this end, a Kinect camera will be used for visual odometry and possibly the BeagleBone's own magnetic compass to correct orientation errors. This sensor fusion approach aims to contribute to the development of a reliable localization system, which is essential for the proposed application and for the safe and efficient interaction of the walker with its users.