1. Locomotion Exploiting Body Dynamics - Semester Project - Student: Matteo de Giacomi Supervisor: Jonas Buchli

2. INTRODUCTION - Purpose of the project - The Puppy II robot - The CPG - Turning

3. Project objectives  Develop a stable and controllable galloping gait for a quadruped robot endowed with passive dynamics  Use of a CPG based on Hopf oscillators

4. Puppy II  4 hip motors  1 spring per knee (passive dynamics)  Sensors (inertia, touch, tortion, IR)  Parameters: Amplitude Frequency Center of rotation

5. CPG  Fully connected system  Matrix describing a galloping in this system: FLFR RLRR

6. Turning  CPG: generates the basic galloping gait  Turn: modifies the basic rythm so that the robot can turn  Actuate:“translates“ the obtained values in values consistent with the robot architecture. CPG Turn Actuate basic rythm feedback Complete behaviour

7. Turning – Setpoint control  Idea: modify the basic position of each leg with a small value FL FR RL RR +Δs+Δs - Δs + Δs - Δs

8. Turning – Amplitude Control  Idea: Increase the amplitude of movement of two ipsilateral legs and decrease the amplitude of their two opposites.

9. PERFORMED TESTS - Introduction - Straight Locomotion - Setpoint Control - Amplitude Control

10. General Framework  Variables influencing PuppyII‘s behaviour: Amplitude Frequency Centers of oscillation  Centers of rotation have been fixed: PuppyII tilted 15° to the front

11. Test 1: Straight Locomotion (1)  Measure of linear speed depending on Amplitude and Frequency  1 measure: space covered over 5 sec  5 measures per test

12. Test 1: Straight Locomotion (2)  Under certain limits in amplitude and frequency, locomotion is stable  Amplitude seems a good way to control the robot‘s speed

13. Videos: Straight Locomotion

14. Tests on Turning Behaviour (1)  Fixed camera 2.45m over the robot  Robot equipped with a red led on its back  Robot behaviour filmed for various parameters  Tracking of the robot (red spot)  Circle estimation in Matlab Estimation of the turning radius of the robot depending on the used parameters

15. Tests on turning behaviour (2)  Example of circle estimation on tracked trajectory

16. Video: Turning

17. Test 2: Setpoint Control  At almost every speed (amplitude) it‘s possible to obtain a good turning behaviour with a good variety of turning radius

18. Test 3: Amplitude Control  At high speed (amplitudes) the turning radius doesn‘t seem to be affected by the used parameter  At low speeds some localized peaks emerge: the robot CAN‘T turn there!

19. CONCLUSION - Discussion - Further works

20. Discussion  Amplitude is a good way to control the robot‘s speed in a range of values contrained by the enviroment and by the robot itself.  Setpoint control is a good way to precisely control the turning radius of the robot  Amplitude control permits large turns at high speeds. At low speed shows a strange behaviour. Feature of the used springs?

21. Further Works  Feedback can improve the gait?  Embed the turning part in the oscillators themself may be useful?  We fixed some parameters (frequency and setpoints). What happens if we change them?

22. THE END Thank you! Any Question?