We analyzed the kinematics of a six-axis robot and explored the possibility of independently predicting its motion trajectory solely based on measured accelerations in three orthogonal directions. The aim of the thesis was to develop a methodology that enables the reconstruction of the robot's motion without direct access to controller data, which we used to calculate the reference trajectory through direct kinematics. The accelerometer was calibrated across a wide frequency range, ensuring accurate signal acquisition and adjusting the sensor's sensitivity. Accelerations were recorded during various robot movements, and the acquired signals were analyzed in the frequency domain using Fourier transform. Special attention was given to noise filtering and the selection of optimal numerical filters and methods. The processed data was used to predict the trajectory, which was compared to the data obtained from the robot's controller. This allowed us to evaluate the accuracy and applicability of the proposed method and identify potential limitations.
|
