Franck Plestan

Absolute Orientation Estimation for Observer-based Control of a Five-link Walking Biped Robot

The design of control laws and reference trajectories for biped robots is difficult, is still a challenging problem and will be not properly solved as long as the dynamics of the robots under interest is not thoroughly understood (see for example some recent papers such as [1]–[10]). Furthermore, the improvement of desired performance and the increasing sophistication of such mechanical systems induce that the complexity of the control, based on the nonlinear model of the biped robots and using all state variables, increases. Accurate sensors are necessary to measure the joint variables, the absolute orientation of the robot and the ground reactions. The determination of the absolute orientation of a biped robot with feet is easy if the sole of the foot has a flat contact with the ground. However, fluid walking gaits induce partial contact of the soles of the feet with the ground: in this case, the biped is underactuated and an accurate measurement of its absolute orientation (with an unstable posture) is, from a technical point-of-view, difficult. Note that it is easier to obtain an accurate measurement of a joint variable equipped with an encoder sensor, especially if this joint is actuated with a motor equipped with a gearbox reducer. Then, there is a real interest to develop observers in order to estimate absolute angular positions and velocities from only the knowledge of the relative angular variables. In order to define the strategy of observation, the biped under interest has point feet and is underactuated in single support: then, its walking gait is dynamically stable.

Sliding Mode Observer with No Orientation Measurement for a Walking Biped

An observer-based controller is proposed to estimate the absolute orientation of a three-link biped without feet, during a cyclic walking gait composed of single supports and impacts. The angular moment is used for the design of the observer, based on second-order sliding mode approach, and for analysis of the stability of the walking.