Robots Learning From Human Motion

robots, technology, motions, u.s.

At the Georgia Institute of Technology, researchers are working on making robots more intelligent – using human motion. By doing this researchers believe they can create robots that work parallel to humans. This is the next development for a world that will one day welcome the use of robotics in everyday life.

Researchers have created arm sensors that are connected to a computer that reads motion. The readings are then analyzed to create a basis from which a robot can learn. Remarkably, when the robot learns the variety and range motions from a human, it can predict and anticipate movements. Most robots in factories are designed to work in a fashion that will not harm human workers. The sensors being developed by Georgia Institute of Technology will be able to provide a safer workplace in the manufacturing industry and will allow robots to complement human labor to a better degree.

Georgia Tech Ph. D. graduate Billy Gallagher stated, “It’s a constant tug of war between the person and the robot.” Both parties react to each other rather than working in unison. The same concept goes to the development of intelligent robotic movement. He explains that the biggest problem in development is that human muscles are never at a constant stiffness or application of force. Robots currently used do not know how to react with variance in hydraulic force.

Jun Ueda, professor at the Woodruff School of Mechanical Engineering and Gallagher’s adviser, said, “The robot becomes confused. It doesn’t know whether the force is purely another command that should be amplified or ‘bounced’ force due to co-contraction.” When force is applied it would create a vibration which becomes increasingly worse as human muscles continue to contract.  Previously, the robot would act regardless; however, that has been changed since the robot can now learn scenarios and the muscle movement associated to them.


The robot is able to learn scenarios by knowing what a human is doing. Researchers believe that dangerous vibrations and forces can be stopped if the human operator wears sensors on their forearms. Doing this would allow the robot to act intelligently and adjust itself and its actions to complement the human. This would also eliminate the dangers associated with intelligent machinery in an industrial setting. Developing these mechanical traits require a better understand of haptic devices that learn from human motion.

Haptic, or non-verbal, means of communicating with robots require physical contact between both operator and machine to create an efficient system. As previously mentioned, operator stiffness is not measurable. By using electromyography (EMG) operator stiffness is able to be quantified in a system allowing for its haptic translation. Thus, stability can be achieved with machines used in many industrial applications that require force-assisted devices.

Ueda states that the team will continue to solve the information translation problem between man and machine. They also received a $1.2 million National Robotics Initiative grant to continue research and development in regards to the mechanisms of neuromotor adaption in human-robot physical interaction. This research will benefit industries with a high interest in advanced manufacturing in the automotive, military, and aerospace technologies. “By making robots smarter, we can make them safer and more efficient,” Ueda said. With the technological developments in the world of robotics being advanced by robots learning from human motion, the future of robotics looks safer and brighter each day.

By: Alex Lemieux




Georgia Tech News Center

You must be logged in to post a comment Login