Jemin Hwangbo
Auto-generated profile, based on publicly available information

Jemin Hwangbo

KAIST

Research Focus & Metrics

Academic Career Length

10Years (2017-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
30 January 2019 in Science Robotics
Verified
Journal Article
Closed access
PDF file
JHJemin Hwangbo
Jemin Hwangbo
JLJoonho Lee
Joonho Lee
ADAlexey Dosovitskiy
Alexey Dosovitskiy
DBDario Bellicoso
Dario Bellicoso
VTVassilios Tsounis
Vassilios Tsounis
MHMarco Hutter
Marco Hutter
A method for learning agile control policies uses simulated data to enable precise, efficient movements in a complex physical robot.

21 October 2020 in Science Robotics
Verified
Journal Article
Closed access
PDF file
JLJoonho Lee
Joonho Lee
JHJemin Hwangbo
Jemin Hwangbo
LWLorenz Wellhausen
Lorenz Wellhausen
VKVladlen Koltun
Vladlen Koltun
MHMarco Hutter
Marco Hutter
A learning-based locomotion controller enables a quadrupedal ANYmal robot to traverse challenging natural environments.

26 January 2022 in Science Robotics
Verified
Journal Article
Closed access
TMTakahiro Miki
Takahiro Miki
JLJoonho Lee
Joonho Lee
JHJemin Hwangbo
Jemin Hwangbo
LWLorenz Wellhausen
Lorenz Wellhausen
VKVladlen Koltun
Vladlen Koltun
MHMarco Hutter
Marco Hutter
Legged robots that can operate autonomously in remote and hazardous environments will greatly increase opportunities for exploration into underexplored areas. Exteroceptive perception is crucial for fast and energy-efficient locomotion: Perceiving the terrain before making contact with it enables planning and adaptation of the gait ahead of time to maintain speed and stability. However, using exteroceptive perception robustly for locomotion has remained a grand challenge in robotics. Snow, vegetation, and water visually appear as obstacles on which the robot cannot step or are missing altogether due to high reflectance. In addition, depth perception can degrade due to difficult lighting, dust, fog, reflective or transparent surfaces, sensor occlusion, and more. For this reason, the most robust and general solutions to legged locomotion to date rely solely on proprioception. This severely limits locomotion speed because the robot has to physically feel out the terrain before adapting its gait accordingly. Here, we present a robust and general solution to integrating exteroceptive and proprioceptive perception for legged locomotion. We leverage an attention-based recurrent encoder that integrates proprioceptive and exteroceptive input. The encoder is trained end to end and learns to seamlessly combine the different perception modalities without resorting to heuristics. The result is a legged locomotion controller with high robustness and speed. The controller was tested in a variety of challenging natural and urban environments over multiple seasons and completed an hour-long hike in the Alps in the time recommended for human hikers.

14 February 2022 in IEEE Robotics and Automation Letters
Verified
Journal Article
Closed access
GJGwanghyeon Ji
Gwanghyeon Ji
JMJuhyeok Mun
Juhyeok Mun
HKHyeongjun Kim
Hyeongjun Kim
JHJemin Hwangbo
Jemin Hwangbo
In this letter, we propose a locomotion training framework where a control policy and a state estimator are trained concurrently. The framework consists of a policy network which outputs the desired joint positions and a state estimation network which outputs estimates of the robot’s states such as the base linear velocity, foot height, and contact probability. We exploit a fast simulation environment to train the networks and the trained networks are transferred to the real robot. The trained policy and state estimator are capable of traversing diverse terrains such as a hill, slippery plate, and bumpy road. We also demonstrate that the learned policy can run at up to 3.75 m/s on normal flat ground and 3.54 m/s on a slippery plate with the coefficient of friction of 0.22.

25 January 2023 in Science Robotics
Verified
Journal Article
Closed access
SCSuyoung Choi
Suyoung Choi
GJGwanghyeon Ji
Gwanghyeon Ji
JPJeongsoo Park
Jeongsoo Park
HKHyeongjun Kim
Hyeongjun Kim
JMJuhyeok Mun
Juhyeok Mun
JHJemin Hwangbo
Jemin Hwangbo
Simulation-based reinforcement learning approaches are leading the next innovations in legged robot control. However, the resulting control policies are still not applicable on soft and deformable terrains, especially at high speed. The primary reason is that reinforcement learning approaches, in general, are not effective beyond the data distribution: The agent cannot perform well in environments that it has not experienced. To this end, we introduce a versatile and computationally efficient granular media model for reinforcement learning. Our model can be parameterized to represent diverse types of terrain from very soft beach sand to hard asphalt. In addition, we introduce an adaptive control architecture that can implicitly identify the terrain properties as the robot feels the terrain. The identified parameters are then used to boost the locomotion performance of the legged robot. We applied our techniques to the Raibo robot, a dynamic quadrupedal robot developed in-house. The trained networks demonstrated high-speed locomotion capabilities on deformable terrains: The robot was able to run on soft beach sand at 3.03 meters per second although the feet were completely buried in the sand during the stance phase. We also demonstrate its ability to generalize to different terrains by presenting running experiments on vinyl tile flooring, athletic track, grass, and a soft air mattress.

12 January 2018 in IEEE Robotics and Automation Letters
Verified
Journal Article
Closed access
JHJemin Hwangbo
Jemin Hwangbo
JLJoonho Lee
Joonho Lee
MHMarco Hutter
Marco Hutter
This letter introduces a new iterative method for contact dynamics problems. The proposed method is based on an efficient bisection method which iterates over each contact. We compared our approach to two existing ones for the same model and found that it is about twice as fast as the existing ones. We also introduce four different robotic simulation experiments and compare the proposed method to the most common contact solver, the projected Gauss-Seidel (PGS) method. We show that, while both methods are very efficient in solving simple problems, the proposed method significantly outperforms the PGS method in more complicated contact scenarios. Simulating one time step of an 18-DOF quadruped robot with multiple contacts took less than 20 μs with a single core of a CPU. This is at least an order of magnitude faster than many other simulators which employ multiple relaxation methods to the major dynamic principles in order to boost their computational speed. The proposed simulation method is also stable at 50 Hz due to its strict adherence to the dynamical principles. Although the accuracy might be compromised at such a low update rate, this means that we can simulate an 18-DOF robot more than thousand times faster than the real time.