|
I am a Robotics Researcher at the Honda Research Institute, where I currently lead and contribute to a project focused on online learning for dexterous manipulation in the real world. I am particularly interested in online planning and data-efficient real-world learning for tasks in which teleoperation demonstrations are not possible and in settings where direct sim-to-real approaches fail.
I received my Ph.D. from Northwestern University in 2024, where I was advised by Kevin Lynch and Jose Pons. During my Ph.D., I worked on interaction force controllers for lower-limb exoskeletons and introduced a novel gait rehabilitation paradigm that enables real-time shaping of gait through bidirectional feedback.
I also hold a robotics-focused master’s degree from ETH Zürich, where I completed my thesis at the Robotic Systems Lab. I have extensive research and hands-on experience with robots that physically interact with their environments.
|
Selected Works
-
Primitive Informed Sampling-based MPC for Multi-fingered Dexterous Manipulation
Honda Research Institute, 2025-2026
TLDR
-
Real-World Reinforcement Learning with Sampling-based MPC Guidance for Multi-fingered Dexterous Manipulation
Honda Research Institute, 2025-2026
TLDR
TLDR:
- Uses sampling-based model predictive control as a structured exploration guide for off-policy reinforcement learning.
- Enables real-world learning with significantly reduced training time, typically 5 to 60 minutes, and very few environment resets.
- Will share more when possible.
-
Therapist-Exoskeleton-Patient Interaction for Gait Therapy
Emek Barış Küçüktabak†, Matthew R. Short†, Lorenzo Vianello†,
Daniel Ludvig, Levi Hargrove, Kevin Lynch, Jose Pons
Accepted for Publication in Science Robotics, 2026
TLDR |
Paper |
Video
TLDR:
- Introduced a novel gait therapy framework, where a therapist and post-stroke patient wear lower-limb exoskeletons virtually connected at the hips and knees through spring-damper elements.
- The bidirectional haptic connection lets the therapist guide the patient's gait while receiving real-time physical feedback about the patient's movement.
- In eight chronic post-stroke participants, TEPI produced larger joint range of motion, step length, and step height than conventional manual therapy while maintaining active participation and high patient motivation.
-
Haptic Transparency and Interaction Force Control for a Lower-Limb Exoskeleton
Emek Barış Küçüktabak, Yue Wen, Sangjoon J Kim, Matthew R Short, Daniel Ludvig,
Levi Hargrove, Eric J Perreault, Kevin Lynch, Jose Pons
IEEE Transactions on Robotics, 2024
TLDR |
Paper |
Video
TLDR:
- Introduced whole-exoskeleton closed-loop compensation (WECC) to estimate interaction torques across the full gait cycle using whole-body dynamics and joint torque measurements.
- Tracks desired interaction torques through constrained optimization, enabling transparent, assistive, or resistive behavior while respecting physical and safety limits.
- Experiments with three subjects showed consistently low torque-tracking error for both zero and nonzero interaction torques, outperforming a simplified dynamic model and passive exoskeleton behavior.
-
Virtual Physical Coupling of Two Lower-Limb Exoskeletons
Emek Barış Küçüktabak, Yue Wen, Matthew R Short, Efe Demirbas,
Kevin Lynch, Jose Pons
IEEE International Conference on Rehabilitation Robotics (ICORR), 2023
TLDR |
Paper |
Video
TLDR:
- Developed a system that renders haptic interaction between two users walking in multi-joint lower-limb exoskeletons by commanding interaction torques from user kinematics and virtual coupling properties.
- Demonstrated soft and hard haptic properties, bidirectional and unidirectional connections, and virtual couplings expressed in both joint space and task space.
- With haptic coupling, dyads generated synchronized movement, and joint-angle differences decreased as virtual stiffness increased.
-
Exoskeleton-Mediated Physical Human-Human Interaction for a Sit-to-Stand Rehabilitation Task
Lorenzo Vianello, Emek Barış Küçüktabak, Matthew Short, Clément Lhoste,
Lorenzo Amato, Kevin Lynch, Jose Pons
IEEE International Conference on Robotics and Automation (ICRA), 2024
(Best Medical Robotics Paper, Best Paper Finalist)
TLDR |
Paper |
Video
-
Towards Dynamic Transparency: Robust Interaction Force Tracking Using Multi-Sensory Control on an Arm Exoskeleton
Yves Zimmermann, Emek Barış Küçüktabak, Farbod Farshidian, Robert Riener, Marco Hutter
ETH Zurich, Master's Thesis, 2018-2019
IEEE International Conference on Intelligent Robots and Systems (IROS), 2020
TLDR |
Master's Thesis |
Paper |
Video
TLDR:
- Introduced a virtual model controller that uses two 6-DoF force sensors, joint accelerations, and inverse dynamics to track interaction wrenches on the multi-DoF torque-controlled ANYexo arm exoskeleton.
- Combined disturbance observation with acceleration estimation from joint encoders and seven IMUs to improve robustness to modeling errors during dynamic human motion.
- Reduced felt inertia and maximum reflected joint torque by more than a factor of three compared with prior transparent controllers, setting a new haptic transparency benchmark for comparable devices.
-
Executing Tasks with a Walking Manipulator: Opening Doors
Emek Barış Küçüktabak.
Advised by Dario Bellicoso, Marko Bjelonic,
Koen Kramer, Marco Hutter
ETH Zurich, Semester Project, 2018
TLDR |
Video
TLDR:
- Developed a door-opening behavior for a walking manipulator, combining a quadruped mobile base with an arm to execute a practical whole-body manipulation task.
- Coordinated base motion and arm motion across the door-opening sequence, including approaching the handle, manipulating the door, and moving through the doorway.
- Evaluated the task with door and manipulator trajectories, demonstrating the feasibility of mobile manipulation on a legged robotic platform.
|
|
