Electromagnetic and inertial motion sensor fusion

Electromagnetic (EM) sensors and inertial sensors (gyros and accelerometers) are two types of motion tracking sensors that could be used as a wearable technology to extract real-time localization values of a teleoperator's controlling appendage and translate them into joint coordinates for a robot arm with the same degrees of freedom. The EM sensor measures time-independent position and orientation, while the gyro and accelerometer measure time-dependent rotational velocity and linear acceleration, respectively. This paper proposes a low-computational dynamically weighted discrete sensor fusion algorithm, based on the complementary filter, to fuse the orientation measurements from these two types of sensor by drawing on the advantages of one to compensate for the flaws of the other one. This algorithm uses the inertial sensor's gyroscopic velocity and the EM sensor's orientation and compares each sensor's higherorder derivatives in order to readjust the filter's weight during each iteration. This allows for the gyro to continuously either validate or correct the EM sensor's orientation. Experimental results using controlled rotations of the sensors performed by an industrial robot show that this new filtration concept can return an accurate orientation value by adequately counteracting the EM sensor's errors from field distortions while also avoiding the inertial sensor's temporal drift.