Search

Browse

• Communities
• Authors
• Year
• Titles
• Subjects

My Space

• Login

Predictive control for stable dynamic locomotion of real humanoid robots

Piperakis Stylianos

Full record

---

Summary

Robust stable omnidirectional locomotion for humanoid robots is a crucial problem and an active research area nowadays. In general, biped locomotion relies on distinct gait phases, during which it must be ensured that the sum of the forces acting on the robot do not result in a loss of balance. To generate stable walking patterns, the need of a stability measure is evident to ensure upright locomotion. State-of-the-art work on this problem uses the Zero Moment Point (ZMP) as a criterion to measure stability. The ZMP approach is a formal representation of the problem, which makes full use of sensor information commonly available on humanoid robots and allows for rigorous solutions to be constructed. This thesis presents a complete formulation of the challenging task of stable humanoid robot omnidirectional walk, based on the Cart and Table model for approximating the robot dynamics. For the control task, two novel approaches are proposed: (i) Preview Control augmented with the inverse system for negotiating strong disturbances and uneven terrain and (ii) Linear Model-Predictive Control (LMPC) approximated by an orthonormal basis for computational efficiency coupled with constraints for improved stability. For the generation of smooth feet trajectory, a new approach based on rigid body interpolation is proposed, enhanced by adaptive step correction. Finally, we present a sensor fusion approach for sensor-based state estimation and an effective solution to sensors' noise, delay, and bias issues, as well as to errors induced by the simplified dynamics and actuation imperfections. The proposed formulation is applied on a real Aldebaran Nao humanoid robot, where it achieves real-time onboard execution and yields smooth and stable gaits.

Available Files

Services