1
1.49 m step-down 18 cm step-up Variable Speed Galloping Control using Vertical Impulse Modulation for Quadruped Robots: Application to MIT Cheetah Robot Hae-Won Park and Sangbae Kim Dept. of Mechanical Engineering, Massachusetts Institute of Technology 1. Objectives Present a controller design scheme which provides a robust and variable speed galloping gait on a quadruped robot. Handle variable speed from slow (~3 m/sec) to fast (~15 m/sec) as well as unexpected variations in ground height. Respect constraints on actuator saturation and ground reaction forces. 3. Our Approach Ground Reaction Force (GRF) Profiles [1][2] 1) Insights from biology 3) Convert to a State Feedback Constant swing time for a wide range of the speed [3] Swing leg retraction Swing Stance Speed (m/sec) Time (Sec) 2) Limit Cycle Design Ground Reaction Force Profiles 60 0 -60 GRFx (N) 350 200 0 GRFy (N) Gait Pattern Swing Leg Retraction Constrained Optimization - EQ: Periodicity condition - INEQ: Satisfy a friction cone, bounded body oscillations, and avoid leg collapsing Minimum of C mt , slip distance, and peak transverse force occurs with a near-zero foot tangential speed [4] . GRF profiles Virtual spring and damper Slip avoidance 2. Control Design Principles Take insights from biology, find a limit cycle, and convert to a state-based control consisting of four phases. Manage transitions between phases using a simple state machine. Gait Pattern Stabilizer State Machine Speed Control via Impulse Modulation Slow Speed Fast Speed Scale 4. Results 1) Speed Control 3 m/s ~ 14.9 m/s (30 Nm torque) Slow speed: Anti-Horse-like Gallop High speed: Cheetah-like Gallop 2) Robustness Test Single Step disturbance Randomly Varying Ground 5. Extensions Obtain trot gait by modifying gait pattern Transition between Trotting and Galloping by smoothly changing gait pattern Robust 3D trotting and galloping (recovery from 150 N sideways push) 6. References [1] Walter et al. (2007) J. of Exp. Biol. [2] Hudson et al. (2012) J. of Exp. Biol. [3] Maes et al. (2008) J. of Exp. Biol. [4] Haberland et al. (2011) IROS [5] Bertram et al. (2009) J. R. Soc. - 80 steps trial on rough ground. - 95% success rate for ground height variation of σ = 17.6 cm Click for video Click for video Click for video Click for video Click for video Click for video Click for video Click for Video O verview

Variable Speed Galloping Control using Vertical Impulse ...parkhw/MITCheetah.pdf · 1.49 m step-down 18 cm step-up Variable Speed Galloping Control using Vertical Impulse Modulation

Embed Size (px)

Citation preview

Page 1: Variable Speed Galloping Control using Vertical Impulse ...parkhw/MITCheetah.pdf · 1.49 m step-down 18 cm step-up Variable Speed Galloping Control using Vertical Impulse Modulation

1.49 m step-down

18 cm step-up

Variable Speed Galloping Control using Vertical Impulse Modulation for Quadruped Robots: Application to MIT Cheetah Robot

Hae-Won Park and Sangbae Kim

Dept. of Mechanical Engineering, Massachusetts Institute of Technology

1. Objectives• Present a controller design scheme which provides a robust and variable speed galloping gait on a quadruped robot.

• Handle variable speed from slow (~3 m/sec) to fast (~15 m/sec) as well as unexpected variations in ground height.

• Respect constraints on actuator saturation and ground reaction forces.

3. Our Approach

• Ground Reaction Force (GRF) Profiles[1][2]

1) Insights from biology 3) Convert to a State Feedback

• Constant swing time for a wide range

of the speed[3]

• Swing leg retraction

Swing

Stance

Speed (m/sec)

Tim

e (

Sec)

2) Limit Cycle Design

• Ground Reaction Force Profiles

60

0

-60GRFx

(N)

350

200

0GRFy

(N)

• Gait Pattern

Swing Leg

Retraction• Constrained Optimization

- EQ: Periodicity condition

- INEQ: Satisfy a friction cone, bounded body

oscillations, and avoid leg collapsing

Minimum of Cmt , slip distance, and peak

transverse force occurs with a near-zero

foot tangential speed[4].

GRF profiles Virtual spring

and damper

Slip avoidance

2. Control Design Principles• Take insights from biology, find a limit cycle, and convert to a state-based control consisting of four phases.

• Manage transitions between phases using a simple state machine.

• Gait Pattern Stabilizer

• State Machine

• Speed Control via Impulse Modulation

Slow Speed Fast Speed

Scale

4. Results1) Speed Control

• 3 m/s ~ 14.9 m/s

(30 Nm torque)

• Slow speed:

Anti-Horse-like Gallop

• High speed:

Cheetah-like Gallop

2) Robustness Test

• Single Step disturbance

• Randomly Varying Ground

5. Extensions• Obtain trot gait by modifying gait pattern

• Transition between Trotting and Galloping

by smoothly changing gait pattern

• Robust 3D trotting and galloping

(recovery from 150 N sideways push)

6. References

[1] Walter et al. (2007) J. of Exp. Biol.

[2] Hudson et al. (2012) J. of Exp. Biol.

[3] Maes et al. (2008) J. of Exp. Biol.

[4] Haberland et al. (2011) IROS

[5] Bertram et al. (2009) J. R. Soc.

- 80 steps trial on rough ground.

- 95% success rate for ground height

variation of σ = 17.6 cm Click for video Click for video

Click for video

Click for video

Click for video

Click for videoClick for video

Click for Video Overview