|| Checking for direct PDF access through Ovid
Legged machines designed to walk on flat or irregular terrain do not need to possess great foot positioning accuracy in order to perform stable motion. However, the propulsion of the body—which normally follows a straight line—, requires feet to follow perfect straight lines—in the body's reference frame—, which must be parallel to each other; otherwise internal foot forces will arise. When this occurs, mainly due to mechanical imperfections and kinematic inaccuracies, foot slippage and changes in the attitude/altitude of the vehicle appear. In the case of legged robots that need to move on rigid structures to achieve stable locomotion, precise foot positioning is required to reduce the foot force interaction. Examples of such robots are legged robots that move by grasping structures or climbing robots that clasp their feet to a wall. Kinematic calibration is a traditional method of improving the accuracy of robot manipulators. These techniques, especially those based on kinematic closed chains, can also be applied to legged robots. This article introduces a method for calibrating legged machines autonomously. The theory has been developed to control a special four-legged robot, and for calibrating purposes each leg has been considered as a 2-DOF leg. Nevertheless, the theory can be easily extended to 3-DOF-leg machines as well as to six-legged machines. The method is of particular interest for industrial machines that walk on rigid structures so that the feet clasp firmly and cannot slip. The method is evaluated through simulation and tested in an industrial four-legged machine developed to walk in a double bottom cell of a ship's hull. Some experiments have been conducted using the calibrated kinematic model to validate the usefulness of the calibration method.