Design and Control of an Autonomous Bipedal Robot

1. Mechatronics Honours Project 2005 School of Mechanical Engineering Stumpy An autonomous bipedal robot Michael Cowling | Andrew Jeffs | Nathan Kaesler Supervisors: Dr Frank Wornle | Mr George Osborne

2. Background 2 Mechanical Engineering • History of bipedal walking robots • 1968~1969: first functional robot: WL-3 • Current: fully functioning humanoidssuch as Honda ASIMO and Sony QRIO • Applications • Prosthetics for the disabled • Entertainment • Human assistance “Bipedal walking robots” www.humanoid.rise.waseda.ac.jp “Multifunctional Above-knee Prosthesis” www.humanoid.rise.waseda.ac.jp Sony 2005 Honda 2005

3. Motivation 3 Mechanical Engineering • University of Adelaide designedpneumatic muscles • Stimulate robotics research at the University

4. Seminar Outline 4 Mechanical Engineering • Project aims • Control techniques • Traditional method: biomechanical • Contemporary method: gait synthesis • Passive dynamic design concept • Design process • Passive biped for downhill walking • Pneumatic muscle actuated biped • Results and future directions

5. Aims 5 Mechanical Engineering • Design and build a unactuated biped • Extend design to incorporate muscle actuation • Make self-contained • Extension goal: incorporate standing still, stopping and starting

6. Design Methodologies 6 Mechanical Engineering • Two main design methods • Biomechanical control • Traditional control method • Robust and versatile • ‘Robotic’ looking gait • Difficult and expensiveto implement • Unnecessarily complex • Inefficient and heavy • Control based on gait synthesis “Asimo X2 at Robodex 2003” http://www.plyojump.com/asimo.html

7. Control by Gait Synthesis 7 Mechanical Engineering • Simulate natural kinematics of walking • Begins with essentials of walking • Actuate only when required • Relatively new approach • McGeer 1990, Wisse 2004 • Inherent sequential design • Suited to muscle actuation • Simple control required • Very efficient Collins et al 2005

8. Control by Gait Synthesis 8 Mechanical Engineering • Passive dynamic concept • Gravitational power only • Perfect starting point • Natural looking gait • Simple and light • Leads to human-likebehaviour • Only dynamically stable Collins et al 2005 Collins et al 2005

9. Preliminary Design 9 Mechanical Engineering • Gait synthesis approach chosen • Simplifications to human physiology • Fewer degrees of freedom • Minimal actuators • Simplest walker concept • Natural starting point • Prototypes for feasibility • ‘Mancano’ promising • ‘Legoman’ unsuccessful ‘Mancano’ “Simplest Walker” http://mms.tudelft.nl/dbl/research/biped ‘Legoman’

10. Design of Stumpy 10 Mechanical Engineering • Extremely difficult task • Discrete events and varying configuration • Non-linear and naturally unstable dynamics • Complex mathematical model • Empirical results required • Design based on McGeer’skneed walking model • Unactuated kneed biped • Made goals achievable McGeer 1990

11. Muscle clamp R Clamp Mass Mechanical Design 11 Mechanical Engineering • Simple mechanical layout • Four legs in pairs • Pinned knee joints • Curved feet • Consideration for tuning and actuation • Limb lengths • Weight distribution • Muscle mounting • Foot position and radius

12. Final Design 12 Mechanical Engineering

13. Testing and Results 13 Mechanical Engineering • Many variables • Limb lengths • Foot radius • Mass distribution • Ramp angle • Starting conditions • Success!

14. Testing and Results 14 Mechanical Engineering • Poor repeatability • Knee bounce main failure mechanism • Knee damping ineffective • Latch required • Very promising for next design stage

15. Biped Actuation 15 Mechanical Engineering • Passive biped has limited functionality • Only walks downhill • Cannot start, stop or stand still • Actuation can overcome these • Modify passive biped • Add actuation • Add associated power and control

16. Valve assembly Bladder Gas supply Power Muscle Actuation 16 Mechanical Engineering • Pneumatic muscle operation • Mains air or bottled CO2 supply (~2 bar) • Power required for switching • Preserves passive dynamic action • Other benefits • Simple construction • Light and powerful • Low cost • Efficient

17. Outer leg lever arm Inner leg lever arm 1 2 3 4 5 Mechanics 17 Mechanical Engineering • Various muscle configurations possible • Actuate knees only • Actuate hips, and use knee latches • Actuate all joints, with either 1 or 2 muscles on each • Five muscles used • Antagonistic hiparrangement • Only two actuators required • Simple lever arm attachment • Muscle-spring systemfor knees

18. Electronics and Control 18 Mechanical Engineering • Motorola 9S12C32 microcontroller on-board • H-Bridge muscle switching • On/off muscle sequencing • Real-time tuning via PC • On-board user controls • Foot switch cycle initialisation • Li-ion batteries for power and electronics Micro H-Bridge

19. Final design 19 Mechanical Engineering

20. Results 20 Mechanical Engineering • Stumpy walks successfully with some assistance

21. Achievements 21 Mechanical Engineering • Stumpy walked passively • Actuated walking successful • Will improve with fine tuning • Self-contained, but with mobile gas supply

22. Future Directions 22 Mechanical Engineering • Standing still, starting and stopping • Turning • Incorporating upper body • More degrees of freedom • Two legs • Active ankles

23. Conclusion 23 Mechanical Engineering • Gait synthesis instead of biomechanical control • Two prototypes built • Passive biped Stumpy based on existing design • Walked successfully • Muscle actuation added to Stumpy • Powered biped walked with some assistance

24. Acknowledgements 24 Mechanical Engineering • Supervisors • Dr Frank Wornle and Mr George Osborne • Mechanical engineering workshop staff • Electronics and instrumentation staff

25. Questions 25 Mechanical Engineering