This paper reports on the bipedal robot Lucy which is actuated by pleated pneumatic artificial muscles. This novel actuator is very suitable to be used in machines which move by means of legs. Besides its high power to weight ratio the actuator has an adaptable passive behavior, meaning the stiffness of the actuator can be changed on-line. This allows to change the natural frequency of the system while controlling angular joint positions. The main control concept intended for Lucy is joint trajectory control while selecting appropriate actuator compliance characteristics in order to reduce control efforts and energy consumption which is of great importance towards the autonomy of legged robots. Presently Lucy has made her first steps with the implementation of basic control strategies.The pleated pneumatic artificial muscle and its characteristics will be discussed briefly and the design of Lucy which is made modular on mechanical as well as electronic hardware level will be described in detail. To pressurize the muscles, a lightweight valve system has been developed which will be presented together with the fundamental control aspects of a joint actuated with two antagonistically setup artificial muscles. Additionally the first experimental results will be shown and briefly discussed.

This paper intends to contribute to the study of dynamically balanced legged robots. A real-time applicable control algorithm for a planar one-legged robot is developed, which allows for locomotion on an irregular terrain. The simulated model consists of an articulated leg and a body, vertically placed upon the leg. During the stance phase the leg is supported by a massless foot. The algorithm is based on the choice of a number of objective locomotion parameters which can be changed from one hop to another. From a chosen initial configuration the robot is able to transfer to a chosen end configuration, while simultaneously controlling its forward velocity, its step length and its stepping height. The foot is thus being placed exactly on a chosen foothold. To reach this goal, the actuators track polynomial functions. The calculation of these functions is based on the objective parameters, and takes into account the constraints acting on the robot. These constraints result from the fact that during flight the center of gravity of the robot tracks a parabolic trajectory, and that the angular momentum with respect to the center of gravity is conserved. Writing the angular momentum constraint in a Caplygin form is the key to the algorithm. Promising simulation results for the algorithm are shown for two different experiments.

This paper reports on the use of a particular actuator in the field of legged robots. The proposed actuator, the Pleated Pneumatic Artificial Muscle, has some interesting characteristics which makes it suitable for machines which move by means of legs. An important issue is the actuator's adaptable passive behaviour which allows the stiffness of a joint that is actuated by two antagonistically coupled muscles, to be varied online. The natural frequency of the system can thus be changed in order to reduce control efforts and energy consumption. The idea of changing this natural frequency in combination with trajectory control will be implemented on a two-dimensional leg model. It will be shown that an appropriate choice of compliance can strongly reduce the amount of needed control activity and energy consumption while tracking a given trajectory.

This paper reports on the control structure of the pneumatic biped "Lucy." The robot is actuated with pleated pneumatic artificial muscles which have interesting characteristics that can be exploited for legged machines. They have a high power-to-weight ratio, an adaptable compliance and they can reduce impact effects. The current control architecture focuses on the trajectory generator and the tracking controller, which is divided into a computed torque controller, a delta-p unit, a PI position controller and a pressure bang-bang controller. The trajectory generator provides polynomial joint trajectories while the computed torque, combined with the delta-p unit, calculates the required muscle pressure levels. The PI and bang-bang controller work at a pressure level to cope with modeling errors and to set the pressures in each muscle. The control design is divided into single support and double support, where specifically the computed torque differs for these two phases. The proposed control architecture is evaluated with a full hybrid dynamic simulation model of the biped. This simulator combines the dynamical behavior of the robot with the thermodynamical effects that take place in the muscle-valves system. The observed hardware limitations of the real robot and expected model errors are taken into account in order to give a realistic qualitative evaluation of the control performance and to test the robustness. A preliminary implementation of the presented controller on the real biped, representing a walking motion of the robot while both feet are in the air, is discussed. This first implementation shows already promising results concerning tracking performance of the proposed control architecture. It confirms that the pneumatic tracking system can be used for a dynamic application such as a biped walking robot.

Summary. A real-time joint trajectory generation strategy for the dynamic walking biped "Lucy" [1, 2] is proposed. This trajectory planner generates dynamically stable motion patterns by using a set of objective locomotion parameters as its input, and by tuning and exploiting the natural upper body dynamics. The latter can be determined and manipulated by using the angular momentum equation. Basically, trajectories for hip and swing foot motion are generated, which guarantee that the objective locomotion parameters attain certain prescribed values. Additionally, the hip trajectories are slightly modified such that the upper body motion is steered naturally, meaning that it requires practically no actuation. This has the advantage that the upper body actuation hardly influences the position of the ZMP. The effectiveness of the developed strategy is proven by simulation results.