Bipedal Locomotion

ME 780 Graduate Course Presentation

Transcript

1. Bipedal Locomotion ME 780: Course Project Safwan Choudhury
2. Biped Locomotion ★ Motivations ★ Project Goals ★ Bipedal Locomotion ★ Control Strategy ★ Next Steps
3. Motivations ★ Navigating Human Environments
4. ZMP Example 2011 Honda Asimo Video
5. Passive Dynamics Passive Dynamic Walker Video
6. MASc Research Goals ★ Virtual Passive Dynamics ★ Foot Placement Estimator ★ Minimizing energy consumption ★ Human-Like Gait Cycles
7. MASc Research Platform
8. MASc Research Platform
9. Project Goals ★ ZMP-Based Walking Controller ★ Sensory Inputs: • Inertial Measurement Units (IMU) • Force/Contact Sensors
10. Zero-Moment Point ★ Point where the inertial and gravity moments cancel out (i.e. the moment is zero). ★ Guarantees stability if this point is within the support polygon.
11. Support Polygon
12. Linear Inverted Pendulum ★ Models entire biped as a single point mass located at the center of gravity (COM)
13. ZMP Equation
14. Contact Model
15. Control Strategy ★ Joint Space Control ★ Operational Space Control ★ Center-of-Mass (COM) Control ★ ZMP-Based Control
16. qref [−0.8 0.3] Kp 0.1 Kd 0.02 ForwardDynamics u q qd qdd Error Signal Command Torque Angular Velocites Angular Positions Joint Space Control
17. Operational Space Control ★ Jacobian matrix provides the mapping between joint space and operational space.
18. Operational Space
19. Operational Space
20. Force Sensors
21. Gravity Compensation q x_com x = Jq qd = inv(Jcom) xd Kp 10 Kd 2 CF qdd qd q u Forward Dynamics u q qd qdd CF qd xd xd qd q com COM Reference (−0.0039 0.0086 0.4022) COM Control
22. Gravity Compensation q x_com x = Jq qd = inv(Jcom) xd x_com x_zmp Stable COM Generator ZMP Reference −C− Kp 10 Kd 2 CF qdd qd q u Forward Dynamics u q qd qdd CF zmp com com zmp qd xd xd qd q com ZMP Control
23. Walking Simulation
24. Walking Simulation
25. Next Steps ★ IMU Modelling + Additive Noise ★ Implementing EKF-Based Estimation ★ State Machine for Cyclic Gait
26. Questions?