|
I will be joining Northeastern University in Jan 2022 as an assistant professor in Mechanical & Industrial Engineering. My lab will focus on modeling the neuromechanics of human movement and applying it to assistive devices and rehabilitation treatment.
I am looking for graduate students and postdocs to join my lab, so please email me if interested in the following directions.
1. Developing versatile gait assistive exoskeleton controllers through neuromechanical simulations
- Keywords: rehabilitation robotics, biomechanics, leg exoskeleton, gait analysis, predictive neuromechanical simulation
- Example tasks
- Design and conduct exoskeleton-assisted gait experiments
- Validate and improve neuromechanical simulation framework
- Develop versatile exoskeleton controllers using simulation framework and validate with human subjects
- Related links:
- NIH K99/R00 project
- Paper 1 (exoskeleton experiment, human-in-the-loop optimization)
- Paper 2 (predictive neuromechanical simulation, elderly gait)
2. Modeling human locomotion control using deep reinforcement learning
- Keywords: motor control, neuromechanical simulation, deep reinforcement learning, open-source software development, academic competition
- Example tasks
- Develop RL frameworks that train physiologically plausible control models
- Develop open-source neuromechanical simulation environments
- Organize academic RL competitions and workshops
- Related links:
- NeurIPS: Learn to Move competition, OpenSim-RL
- Paper 1 (neuromechanical simulation, deep RL)
- Paper 2 (physiologically plausible control model, neuromechanical simulation)
For the postdoc positions, I am looking for researchers with one or more of the following expertise:
- Musculoskeletal simulation for human motor control
- Human gait experiment/analysis for assistive device
- Deep reinforcement learning for continuous control
- Physics simulation for legged characters
I am a postdoctoral fellow at the Mechanical Engineering Department of Stanford University. I work with Drs. Steve Collins on gait assistive exoskeletons. I am developing human-in-the-loop optimization frameworks for exoskeleton assistance that can improve human locomotion performance in all dimensions of comfort, safety, efficiency, and speed. I am also interested in better understanding motor learning and across-subject variability we observe from subjects walking in exoskeletons.
I am a recipient of the NIH Pathway to Independence Award (K99/R00, NIA) for the project Simulation framework to develop ankle exoskeleton gait assistance for older adults. In this project, I will adapt the neuromechanical models from my previous studies to predict outcomes of exoskeleton assistance.
I am also the lead organizer of the NeurIPS 2019: Learn to Move competition. Through this competition, we aim to bridge researchers from neuroscience, biomechanics, robotics, and machine learning to model neuro-musculo-skeletal dynamics of human movement.
- Competition page: https://www.aicrowd.com/challenges/neurips-2019-learn-to-move-walk-around
- Project page: http://osim-rl.stanford.edu/
- Github repo: https://github.com/stanfordnmbl/osim-rl
My broader goal is to better understand and model the motor control and biomechanics of human locomotion and to apply the obtained knowledge to rehabilitation engineering. I develop computational models and conduct human experiments to propose and evaluate control hypotheses. For my Ph.D. study at the Robotics Institute of Carnegie Mellon University with Dr. Hartmut Geyer, I developed and evaluated neuromechanical control models to understand the circuitry-to-behavior layer of human locomotion. I envision my research to bring about simulation testbeds (which I call the digital locomotor clones) for training protocols and assistive devices as well as intelligent motion controllers for assistive and humanoid robots.
Journal papers
|
Optimized hip-knee-ankle exoskeleton assistance reduces the metabolic cost of walking with worn loads GM Bryan, PW Franks, S Song, R Reyes, MP O'Donovan, KN Gregorczyk & SH Collins submitted [preprint] |
|
Optimized hip-knee-ankle exoskeleton assistance at a range of walking speeds GM Bryan, PW Franks, S Song, AS Voloshina, R Reyes, MP O'Donovan, KN Gregorczyk & SH Collins submitted [preprint] |
|
Deep reinforcement learning for modeling human locomotion in neuromechanical simulation S Song, Ł Kidziński, XB Peng, C Ong, J Hicks, S Levine, CG Atkeson & SL Delp Journal of NeuroEngineering and Rehabilitation, 2021 [paper] [video 2017] [video 2018] [video 2019] |
|
Optimizing exoskeleton assistance for faster self-selected walking S Song & SH Collins IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2021 [paper] [video 1] [video 2] [video Stanford] [press] |
|
Using force data to self-pace an instrumented treadmill and measure self-selected walking speed S Song, H Choi & SH Collins Journal of NeuroEngineering and Rehabilitation, 2020 [paper] [code] [video] |
|
Predictive neuromechanical simulations indicate why walking performance declines with aging S Song & H Geyer The Journal of Physiology, 2018 [paper] [press] |
|
Evaluation of a Neuromechanical Walking Control Model Using Disturbance Experiments S Song & H Geyer Frontiers in Computational Neuroscience, 2017 [paper] [video 1] [video 2] |
|
A neural circuitry that emphasizes spinal feedback generates diverse behaviours of human locomotion S Song & H Geyer The Journal of Physiology, 2015 [paper] [video] [model] [journal cover] |
Conference papers
|
Bayesian optimization using domain knowledge on the ATRIAS biped A Rai, R Antonova, S Song, W Martin, H Geyer & CG Atkeson IEEE ICRA, 2018 [paper] [video] |
|
Towards a hierarchical neuromuscular control model with reflex-based spinal control
- A study with a simple running model S Song International Symposium on Advanced Intelligent Systems, 2015 [paper] [video] [best presentation award] |
|
Regulating speed in a neuromuscular human running model S Song & Hartmut Geyer IEEE Humanoids, 2015 [paper] [video] |
|
Toward a virtual neuromuscular control for robust walking in bipedal robots Z Batts, S Song & Hartmut Geyer IEEE IROS, 2015 [paper] [video] |
|
Development of a bipedal robot that walks like an animation character S Song, J Kim & K Yamane IEEE ICRA, 2015 [paper] [video] [talk] [press] |
|
Integration of an adaptive swing control into a neuromuscular human walking model S Song, R Desai & H Geyer IEEE EMBC, 2013 [paper] [video] |
|
Generalization of a muscle-reflex control model to 3D walking S Song & H Geyer IEEE EMBC, 2013 [paper] |
|
The effect of foot compliance encoded in the windlass mechanism on the energetics of human walking S Song, C LaMontagna, SH Collins & H Geyer IEEE EMBC, 2013 [paper] [video] [project page] |
|
Regulating speed and generating large speed transitions in a neuromuscular human walking model S Song & H Geyer IEEE ICRA, 2012 [paper] [video] [slides] |
|
The energetic cost of adaptive feet in walking S Song & H Geyer IEEE ROBIO, 2011 [paper] [slides] [project page] |
|
Development of an omnidirectional walking engine for full-sized lightweight humanoid robots S Song, YJ Ryoo & DW Hong ASME IDETC, 2011 [paper] [video1] [video2] [slides] [project page] |
|
Linear-time encodable rate-compatible punctured LDPC codes with low error floors S Song, D Hwang, S Seo & J Ha IEEE VTC, 2008 [paper] |
Conference abstracts
S Song, H Choi, K Poggensee, CG Atkeson & SH Collins, Human-in-the-loop optimization of ankle-exoskeleton assistance for faster preferred walking speed: a preliminary study, Dynamic Walking, 2019.
[5 min talk]
S Song, Ł Kidziński, R Khidorka, C Ong, S Mohanty, J Hicks, J Ku, S Carroll, S Levine, M Salathé, CG Atkeson, SH Collins & S Delp, Learn to Move: a competition to bridge biomechanics, neuroscience, robotics, and machine learning to model human motor control, Dynamic Walking, 2019.
[10 min talk + Q&A]
S Song, H Geyer, SH Collins & CG Atkeson, Towards predictive neuromechanical simulations for pathological gait and assistive devices, World Congress of Biomechanics, 2018.
[poster]
A Falisse, G Serrancoli, C Dembia, S Song, I Jonkers & F De Groote, "Computationally efficient pre-dictive muscle-driven simulations of 3D walking," World Congress of Biomechanics, 2018.
S Song, Y Aucie & G Torres-Oviedo, Can split-belt treadmill walking be explained with a reflex-based model?, Neuroscience, 2017.
[poster]
S Song & H Geyer, Modeling and exploring elderly walking with neuromechanical simulations, Dynamic Walking, 2017.
[12 min talk + Q&A]
S Song & H Geyer, A spinal reflex based neuromuscular model of human locomotion investigated against unexpected disturbances, Neuroscience, 2016.
[poster]
S Song & H Geyer, Testing a neuromuscular locomotion control model against human experiments, Dynamic Walking, 2016.
[10 min talk + Q&A]
S Song & H Geyer, Using a neuromuscular model of human locomotion to control bipedal robots, Dynamic Walking, 2015.
[poster]
S Song & H Geyer, Robust 3D locomotion models using primarily reflex control, Dynamic Walking, 2013.
[10 min talk + 10 min Q&A]
Patents
J Kim, K Yamane & S Song, Method for developing and controlling a robot to have movements match-ing an animation character, United States Patent 9427868, 2016.
[link]
J Nam, J An, D Hwang, J Ha & S Song, Apparatus and method for encoding low density parity check code, Korean patent 10-0999272-00-00, 2010.
[link]
Invited talks
BioRob 2020 Workshop on Community-Based Rehabilitation Research using Wearable Devices, November 2020.
“Toward predictive simulation for rehabilitation treatment”
[15 min talk + 5 min Q&A]
University of Delaware, September 2020.
“Towards predictive simulation of human movement for assistive devices and rehabilitation treatment”
[40 min talk + 20 min Q&A]
abstract
WearRAcon, March 2020.
“Towards optimal gait assistance”
[20 min talk + 5 min Q&A]
[link]
NeurIPS Deep RL workshop, December 2019.
“Learn to Move competition”
[15 min talk]
[link]
Universities in Europe, July 2018
“Modeling human locomotion control and its applications”
[40 min talk + 20 min Q&A]
- EPFL, Switzerland
- University of Tübingen, Germany
- University of Stuttgart, Germany
- Heidelberg University, Germany
- TU Darmstadt, Germany
- KU Leuven, Belgium
- University of Twente, Netherlands
Universities and research institutes in Korea, July 2017
“Neuromechanical control models of human locomotion and their applications to rehabilitation engineering”
[40 min talk + 20 min Q&A]
- Seoul National University
- Korea Institute of Industrial Technology
- Pohang University of Science and Technology
- Korea Institute of Machinery and Materials
- Inha University
Universities and companies in Korea, November 2015
“A neuromuscular model of human locomotion and its applicaiton to robotic assistive devices”
[40 min talk + 20 min Q&A]
- Chung-Ang University
- Samsung Advanced Institute of Technology
- KAIST
- ROBOTIS
- Seoul National University
The 10th workshop on humanoid soccer robots at IEEE Humanoids, November 2015.
“Neuromuscular model of human locomotion that can generate diverse locomotion behaviors”
[30 min talk]
[link]
Technical writings in Korean
“Understanding the control of human locomotion through simulation and its application to robotic assistive devices,” MATERIC (mechanical and construction engineering research information center), February, 2016. [article]
“Robotic lower-limb prosthetics related technical issues - 2. Control,” ROBOT (monthly magazine), May 2013.
[article]
“Robotic lower-limb prosthetics related technical issues - 1. Hardware,” ROBOT (monthly magazine), April 2013.
[article]