A New Kinematic Kalman Filter (KKF) for End-Effector Sensing of Robotic Manipulators

Abstract

The kinematic Kalman filter (KKF) is a sensor-based state estimator which is immune to the external disturbances and the parameter uncertainties of mechanical plants. This paper extends the original idea of the KKF to a more general form as a means to enhance a real time vision sensor for the end-effector control of a robot manipulator, the performance of which is often limited by its slow sampling rate. The original one-dimensional KKF is reformulated in a higher dimensional form by incorporating the measurements from the vision sensor, accelerometers and gyroscopes. A nonlinear state space model of the kinematics of the end-effector is derived including the time delay associated with vision sensing. Then, the new KKF is formulated as a state estimator combining the inter-sample predictions with an extended Kalman filter (EKF). The paper discusses practical issues such as the real time computation to implement the EKF and the vision sensor to measure the absolute position. Experimental results are presented to confirm the benefits of the new KKF using a two-link direct drive manipulator equipped with a dual axis MEMS accelerometer, a single axis MEMS gyroscope and an end-effector mounted vision camera. The accurate estimation of the position and velocity of the end-effector from the new KKF will be useful for the real time visual-servo and the task space control of robot manipulators.