Rolling! Energy Control Concepts for

Humanoid Falls

Jinoh Lee

October 1, 2018

Advanced Robotics Department

Istituto Italiano di Tecnologia (IIT)

Genova, Italy

Half-day Workshop

Humanoid Robot Falling:

Fall Detection, Damage Prevention, and Recovery Actions

Researcher

Topics

• Controlled fall problem in humanoids

– Related research

– Energy control concepts for humanoids fall

– Challenges and open issues

Rajesh Subburaman, Nikos Tsagarakis, Jinoh Lee, "Online Rolling Motion Generation for Humanoid Falls Based on

Active Energy Control Concepts", 2018 IEEE-RAS International Conference on Humanoid Robots (Humanoids 18),

Nov. 6-9, Beijing, China, (Accepted)

Rajesh Subburaman, Jinoh Lee, Darwin G. Caldwell, Nikos Tsagarakis, “Online Falling-Over Control of Humanoids

Exploiting Energy Shaping and Distribution Methods”, 2018 IEEE International Conference on Robotics and

Automation (ICRA 2018), pp448-454, Brisbane, Australia, May 21-25, 2018

Humanoids Falling over

• Two different fall cases

• What is ‘good falling’..?

• Controlled fall

WALK-MAN (IIT) in DRC FINAL

• Pre-planning falling sequence– Fujiwara, K., Kanehiro, F., Kajita, S., Kaneko, K., Yokoi, K. and Hirukawa, H., 2002. “UKEMI:

Falling motion control to minimize damage to biped humanoid robot.” IROS 2002

– Fujiwara, K. et. al., “The first human-size humanoid that can fall over safely and stand-up

again”. IROS 2003.

– Fujiwara, K. et. al., “Falling motion control of a humanoid robot trained by virtual

supplementary tests”, ICRA 2004.

– Wilken, T., Missura, M. and Behnke, S., “Designing falling motions for a humanoid soccer

goalie.”In Proceedings of the 4th Workshop on Humanoid Soccer Robots (Humanoids 2009)

Research on controlled fall of humanoids

Fujiwara et al. 2002 Fujiwara et al. 2004 Wilken et al. 2009

• Realtime falling motion generation– Fujiwara, Kiyoshi, et al. "An optimal planning of falling motions of a humanoid robot." IROS 2007.

– Ogata, Kunihiro, Koji Terada, and Yasuo Kuniyoshi. "Real-time selection and generation of fall damage

reduction actions for humanoid robots." Humanoid Robots, 2008

– Lee, S.H. and Goswami, A., “Damage minimization of humanoid robots by falling on targeted body

segments”, Journal of Computational and Nonlinear Dynamics, 2013

– Goswami, A., Yun, S.K., Nagarajan, U., Lee, S.H., Yin, K. and Kalyanakrishnan, S., “Direction-changing fall

control of humanoid robots: theory and experiments” Autonomous Robots, 2014.

– Li, Qingqing, Xuechao Chen, Yuhang Zhou, Zhangguo Yu, Weimin Zhang, and Qiang Huang. "A minimized

falling damage method for humanoid robots." International Journal of Advanced Robotic Systems, 2017

– Li, Qingqing, et al. "A Falling Forwards Protection Strategy for Humanoid Robots." ROMANSY 22–Robot

Design, Dynamics and Control., 2019

Research on controlled fall of humanoids

Fujiwara et al. 2007 Lee and Goswami 2013

Li et al. 2017

• Realtime falling motion generationwith active compliance control– Samy, Vincent, and Abderrahmane Kheddar. "Falls control using posture reshaping and active

compliance." Humanoids 2015

– Samy, Vincent, Karim Bouyarmane, and Abderrahmane Kheddar. "Qp-based adaptive-gains

compliance control in humanoid falls." ICRA 2017

– Cardona, Gustavo A., et al. "Reduction of impact force in falling robots using variable

stiffness." SoutheastCon, 2016.

– Luo, Dingsheng, et al. "Biped robot falling motion control with human-inspired active

compliance." IROS, 2016

Research on controlled fall of humanoids

Vincent et al. 2015 Vincent et al. 2017

• Learning based algorithms– Ruiz-del-Solar, J., Palma-Amestoy, R., Marchant, R., Parra-Tsunekawa, I., & Zegers, P. “Learning to

fall: Designing low damage fall sequences for humanoid soccer robots”. Robotics and

Autonomous Systems, 2009

– .Ha, Sehoon, and C. Karen Liu. "Multiple contact planning for minimizing damage of

humanoid falls.“IROS 2015

– Kumar, Visak CV, Sehoon Ha, and C. Karen Liu. "Learning a unified control policy for safe

falling." IROS 2017.

Research on controlled fall of humanoids

Ruiz-del-Solar et al. 2009

Sehoon et al., 2015

Breakfall techniques

• We (humans) need to be trained and learned!

• Problem definition:

How to generate whole-body motion during the fall-

over exploiting breakfall techniques?

Controlled fall strategy for humanoids

Lowering body height

Minimize the total energy of the

system

Rolling action

Select Multiple contact points

Distribute the minimized energy

• Problem definition:

How to generate whole-body motion during the fall-

over exploiting breakfall techniques?

Controlled fall strategy for humanoids

+ =

Energy shaping

(ES)

Energy distribution polygon

(EDP) Controlled fall

+

Energy control concepts

Energy Shaping (ES) control

Telescopic inverted pendulum (TIP) model

In free fall:

Potential energy (EP) is

completely converted to

kinetic energy (EK), resulting

in high impact

0

To be minimized

Shaping the total energy

• Create a CoG control command by an energy control law

Energy Shaping (ES) control

Control input

Energy

Control Law

TIP

0

• Energy Control Law design

• TIP dynamics

• Resulting desired dynamics

• Controlled fall

– energy shaping minimizes the total energy of the system (ET),

lowering impact forces.

• Free fall

(uncontrolled)

ET

ET

Energy Shaping (ES) control

• Form a polygon by selecting multiple contact points to distribute the

energy from impacts

• modify the EDP:

– Hand control H Hnew to make a contact as early as possible.

Energy Distribution Polygon (EDP)

Forward Fall Rightside Fall

Hnew

Hnew

K

TfwdTbwd

Hnew

S

C

Initial EDP

Tside

Backward fall

Whole-body control with ES + EDP

Task prioritization

QP-based optimization

(Open SOT library)

• Front fall case

Whole-body control with ES + EDP

Only ES ES + EDP

Controlled fall

Whole-body control with ES + EDP

Thank you for your attention

• Call for Contribution: Short talk / Posters

Darwin Caldwell,

ADVR DirectorNikos Tsagarakis,

HHCM PIRajesh Subburaman

PhD Student

Workshop in IEEE/-RAS Humanoids 2018

November 6, 2018

Chris Atkeson, Carnegie Mellon University

Joohyung Kim, Disney Research

Jinoh Lee, Istituto Italiano di Tecnologia

Katsu Yamane, Honda Research Institute USA

Alex Alspach, Toyota Research Institute