State Estimation with Static Displacement Compensation for Large-Scale Manipulators

Abstract

Tasks of large-scale manipulators on construction sites are usually planned for the tool center point in absolute coordinates. Automatic tool center point control requires an accurate position feedback either from measurements or from an estimator. Equipping mass-produced construction machines with high-precision and redundant sensors is expensive. Thus, a state estimator based on only one sensor for each joint is developed as an alternative in this work. Using a prediction step, a Kalman filter smooths the sensor measurements without creating a phase delay yielding estimated joint positions and velocities, from which the tool center point position is then estimated using forward kinematics. A static displacement model is derived based on the Euler-Bernoulli beam equation to account for not captured boom deflections caused by the flexible manipulator structure and the effect of gravity. This cheap, computationally light, and easy approach results in a position estimate for large-scale manipulators with an accuracy of a few centimeters throughout the whole task space and a maximum relative error of 2.53 % w.r.t. the manipulator outreach.