[Researcher of the Month] Simulation of Human Movements
Professor Kwon Tae-soo (Department of Computer Science & Engineering)
|Copy URL / Share SNS||
Human movements are much more intricate and complicated than it seems. Many attempts were done to portray moving human actions by computer program and animation. Those attempts were partly successful until now, yet with certain limitations. Professor Kwon Tae-soo of the Department of Computer Science & Engineering is greatly interested in simulating human motion. In his recent paper, “Momentum-Mapped Inverted Pendulum Models for Controlling Dynamic Human Motions”, he explains about how physics can be applied into animating human movement and be used in its development.
Simulating human animation is a complicated business. Most of the animation we see in games and movies is based on a technical method called motion capture. Motion capture is a method of simulating motions by attaching sensors to a moving object and tracking the information of the movements, then analyzing its numerical data.
However, movements of these animations have certain limits because of its foundation which merely consists of pre-captured motions. Therefore, in order to exceed this disadvantage, quite a few research was done utilizing physics into developing animation using Inverted Pendulum Model, or IPM, which analyzes human motions through controlling robots by computerized robot simulator program.
Although IPM became a potentially alternative method of producing simulation of motions, it had a problem of producing unnatural movements of characters. Kwon, who was aware with this limitation of IPM, developed a new form of IPM called Momentum-Mapped Inverted Pendulum Models (MMIPM).
The similarity of IPM and MMIPM is that both methods use two kinds of robot, a simple kind of robot, an upside-down kind of pendulum which is comprised of a cart and a pole, and a humanoid. Due to the difficulty of controlling a complex humanoid, the simple robot is first used. By using conversion after mapping the present state of simple robot, signals for controlling the humanoid can be calculated.
One of the main contrasts between IPM and MMIPM is the way mapping is done. While mapping for IPM must use both the center of mass and center of pressure of the robot for mathmatical differentiation, momentum-mapping uses the center of mass. Differentiating one time instead of two is highly beneficial because the quality of signals improve. In addition, if two feet of the humanoid are above the ground, center of pressure becomes absent, mapping with conventional IPM method become impossible, whereas mapping with MMIPM is still possible.
MMIPM also concentrates on modeling the changes of postures and how much the human body is tilted during performing certain actions. Therefore, because of the differences or technical improvements of MMIPM compared with IPM, expressing more natural and difficult movements can be realized. As a result, Kwon could successfully produce more natural movements of running, and complex acrobatic motions such as spinning, backflip, and handstand.
Professor Kwon’s future studies also focus on human movements, which are reenacting motions of soft parts of the human body, such as fat. According to Kwon, the technology which is used for today’s animations and games is from a decade ago. “Although at first a game with great graphics may seem like something big. However, when you start an online game, soon you will realize that the actions of your characters are mere repetitive movements, ” said Kwon. Through his study, Kwon aspires to broaden the limits of present day game-play and animation. “My ultimate objective is enabling game characters to perform unexpected movements when players enjoy unpredictable game plays,” Kwon revealed.
Jang Soo-hyun email@example.com
Photos by Choi Min-ju
This week's top news
Korean Traditional Colors
[Excellent R&D] Laying the Stepping Stones for Future Software Technology
[Researcher of the Month] Finding a Cure Through Direct Intranasal Delivery
Hanyang University Hospital Receives Top Ranking
[Excellent R&D] Organic-Inorganic Hybrid Multi Layers
Hanyang University Ranks 151st in 2018 QS World University Rankings
Selected for Excellence in University Evaluation on the Perspective of Industry
[Researcher of the Month] Efficient and Aesthetic Hybrid Solar Cells
Structural Roles of gRNAs in the CRISPR-Cas9 System