We present a framework for robust estimation of the configuration of an articulated robot using a large number of redundant proprioceptive sensors (encoders, gyros, accelerometers) distributed throughout the robot. Our method uses an Unscented Kalman Filter (UKF) to fuse the robot’s sensor measurements. The filter estimates the angle of each joint of the robot, enabling the accurate estimation of the robot’s kinematics even if not all modules report sensor readings. Additionally, a novel outlier detection method allows the the filter to be robust to corrupted accelerometer and gyro data.
- Dynamic Systems and Control Division
Robust State Estimation With Redundant Proprioceptive Sensors
Rollinson, D, Choset, H, & Tully, S. "Robust State Estimation With Redundant Proprioceptive Sensors." Proceedings of the ASME 2013 Dynamic Systems and Control Conference. Volume 3: Nonlinear Estimation and Control; Optimization and Optimal Control; Piezoelectric Actuation and Nanoscale Control; Robotics and Manipulators; Sensing; System Identification (Estimation for Automotive Applications, Modeling, Therapeutic Control in Bio-Systems); Variable Structure/Sliding-Mode Control; Vehicles and Human Robotics; Vehicle Dynamics and Control; Vehicle Path Planning and Collision Avoidance; Vibrational and Mechanical Systems; Wind Energy Systems and Control. Palo Alto, California, USA. October 21–23, 2013. V003T40A005. ASME. https://doi.org/10.1115/DSCC2013-3873
Download citation file: