-
Mateus Pereira
Programa de Pós-Graduação em Engenharia Elétrica, Universidade Federal de São Carlos, São Carlos, SP
-
Edson Francelino
Programa de Pós-Graduação em Engenharia Elétrica, Universidade São Paulo, São Carlos, SP
-
Lucca B. Castro
Programa de Pós-Graduação em Engenharia Elétrica, Universidade Federal de São Carlos, São Carlos, SP
-
Roberto S. Inoue
Departamento de Computação, Universidade Federal de São Carlos, São Carlos, SP
-
Samuel Nogueira
Programa de Pós-Graduação em Engenharia Elétrica, Universidade Federal de São Carlos, São Carlos, SP
Keywords:
Self-aligning sensors, unstable fixed points, Extended Kalman Filter with Constraints, Complementary Kalman Filter, exoskeleton, non-linear systems, sensor fusion
Abstract
This paper proposes the use of consecutive inertial sensors to estimate joint angles in the human body, as well as in rigid devices such as exoskeletons and articulated robots, consid- ering the occurrence of possible online misalignments and automatic adaptation, promoting the system’s robustness to such uncertainties. The proposal consists of modeling the problem as a spherical joint with projection of movements in the plane of greater angular amplitude, defining a minimization problem for the automatic virtual alignment of the inertial sensors. In the study, two filtering systems based on Kalman were developed to achieve the proposal, the first being used as a tool for online optimization and adjustment of the virtual alignment parameters of the sensors, while the second was responsible for the fusion of the signals in order to estimate the joint angle of the molded joint. The results obtained were promising, with a Pearson correlation of around 0.99 and RMSE close to 1o when compared with measurements from an encoder, and the system adaptation to sudden changes in orientation took a few seconds after the occurrence, this being dependent on the size of the fitting data window.
