Robot, do what I do
New teleoperation system promises big improvements in humanoid robotics.
Video of teleoperation of stepping in place.
CREDIT: Ramos and Kim, Sci. Robot. 4, eaav4282 (2019)
By Ian Connellan
While improvements in artificial intelligence continue apace, the development of humanoid robots capable of matching human-level movement and dexterity in real situations remains more hope than reality.
Such robots would be ideal first responders to dangerous natural or human-made disasters; they could take the place of a human at a malfunctioning nuclear power plant, for instance.
But in order to successfully handle such situations, robots would have to be capable of doing what human emergency services worker can do. They’d have to be able to navigate unstructured and possibly unstable terrain, and to have the dexterity to use devices designed for human workers.
According to a study published in the journal Science Robotics, we’re now a few steps closer to robots that fit the bill.
US-based engineers Joao Ramos, from the University of Illinois, and Sangbae Kim, from Massachusetts Institute of Technology, have created a new teleoperation (remote control) system that more seamlessly transfers the movement of a human operator to a two-legged robot.
Many existing teleoperation approaches have involved the time-consuming process of capturing human data and manipulating it with computers in an attempt to match a robot’s physical structure and limitations. However, a human operating a robot in this way doesn’t get any physical feedback on what the robot is doing – which would allow for more accurate motion transfer.
Ramos and Kim tackled these challenges using a control strategy that dynamically synchronised a human operator’s motion to that of a small bipedal test robot named Little HERMES.
To better scale human motion to the robot, the researchers used a simplified model for two-legged dynamics called linear inverted pendulum (LIP). With this model, the control system generated feedback forces to the operator proportional to the relative speed between human and robot.
For example, the system would speed up human motion to match a faster robot or generate drag to match the operator to a slower robot.
Ramos and Kim hope to improve their teleoperation system by using more advanced two-legged robots, minimising communication delay between operator and robot, and exploring other ways human intention can be anticipated: through biosignals, for example.
The introduction to their paper dramatically illustrates the worth of this work: “If this technology had been available back in March 2011, the catastrophic outcome of the Fukushima Daiichi power plant nuclear disaster could have been vastly mitigated.
“It is estimated that, if a responder had been able to endure the deadly levels of radiation and enter the facility within the first 24 hours after the cooling system malfunctioned, the first nuclear reactor could have been stabilised.”