KAIST (Korea Advanced Institute of Science and Technology) is going to incorporate AI (Artificial Intelligence) into humanoid robot called ‘Hubo, which represents South Korea, and is going to greatly build up its performance in movement.
KAIST is going to use total of $13.0 million (15.0 billion KRW) for next 5 years in order to upgrade performance of Hubo’s hardware and add AI functions that allow Hubo to learn repeatedly by utilizing Deep Learning technologies with Korea Electrotechnology Research Institute and Kookmin University.
“Ministry of Trade, Industry and Energy has decided to support us with $13.0 million (15.0 billion KRW) for next 5 years.” said Manager Oh Jun-ho of KAIST Humanoid Robot Research Center. “We are planning to develop an upgrade version of ‘Hubo’ that has its performance greatly improved.”
Hubo is a humanoid robot that a research team led by Manager Oh has developed since 2002. After making body (KHR-1) of Hubo in August of 2002, it now has developed ‘DRC Hubo II’ version. It won first prize of DRC (DARPA Robotics Challenge) in 2015 by completing 8 tasks such as turning off valves and jumping over obstacles in fastest time at 44 minutes and 28 seconds.
Research team is planning to have Hubo perform mixed tasks with just one command by incorporating AI into Hubo. Until now, Hubo was only able to perform tasks by programming individual commands into Hubo. Research team had to input conditions such as destination and direction consistently into Hubo in order to move Hubo to a particular location.
Research team is planning to get rid of these inconveniences by actively utilizing Deep Learning technologies. It is going to have Hubo seek solutions on its own by having it repeatedly learn information on environment that can act as a thing or an obstacle. Its first goal is to have Hubo perform tasks from last year’s DRC competition without any detailed commands.
Research team is also going to improve performance of Hubo’s hardware. It is going to improve speed of its movement by increasing power such as instant power that is located around Hubo‘s knees by two to three times. It is planning to apply hydraulic driving mechanism with strong power because current Hubo’s strength is weak as it operates under electric driving mechanism.
Research team is also going to work on securing safety and credibility at the same time. It is going to remove frequent malfunctions by miniaturizing, simplifying, and modularizing components and have Hubo not fall over by responding to slip and steep slope within just 0.3 second. It is planning to make its software (operating system_ with open source.
Goal of KAIST is to upgrade its Hubo as an AI robot that is more outstanding than ‘Atlas’ from the U.S. Atlas, which is a humanoid robot developed by Dynamics from Boston, has become a hot topic as it is able to walk fast even on snow-covered fields and steep slope and stand up rapidly even when it falls over.
“Humanoid robot is a field that advanced countries research on with their pride on the line and our goal is to make a robot that can perform tasks and safely and freely walk on rugged terrain.” said Manager Oh. “We are going to show notable results within two years.”
Staff Reporter Kim, Youngjun | kyj85@etnews.com