Motion generation and motion control are among the key issues in mobile robotics research. Both are highly affected by the type of locomotion of the mobile robot. Due to their energy efficient motion Wheeled Mobile Robots (WMR) are predominant on flat terrain. However, on soft ground WMR loose their efficiency. Hence, in impracticable terrain Legged Mobile Robots (LMR) are more suitable. For technical applications quadruped LMR are of particular interest as four is the minimum number of legs necessary for static walking. In this paper a generic free gait algorithm for motion generation of quadruped robots is discussed. The application of the free gait algorithm on a particular walking robot demands lower level control of the actuated legs. In the case of large scale walking machines as for example ALDURO (weight -1.8t), hydraulic actuators are beneficial due to their power density. Each of the 16 degrees of freedom of ALDURO is actuated by a differential hydraulic cylinder. Therein, sensors provide the pressures in all chambers of the cylinders as well as the positions of all pistons. However, in order to control also the velocity of each cylinder more information is necessary. For this reason a PI-observer is proposed to estimate the velocity of the pistons and the external loads on the cylinders. The load on the actuators is a crucial information for the higher level motion generation, e.g. to determine which leg should be lifted next. Finally, simulation results demonstrate the performance of the observer concept.