195 — Brain-Computer Interface meets ROS: A robotic approach to mentally drive telepresence robots

Beraldo et al (1712.01772)

Read on 03 March 2018
#non-invasive  #BCI  #EEG  #ROS  #robotics  #telepresence  #drive  #navigation 

Pepper is a humanoid wheeled robot. The researchers chose to adapt Pepper to drive based upon commands from a driver (rather than in its autonomous mode) — and they chose to drive the robot using thought.

Using EEG signals from a “driver”, the software converts “left” and “right” instructions (where the human imagines moving both hands and feet on the respective side of their own body) into instructions to pass to the ROS robotic control system. The signals are parsed using a weighted sliding-window average of the driver’s interpreted commands.

Because this system leverages ROS, the researchers were also able to incorporate other ROS-based navigation systems, such as obstacle avoidance and domain mapping. Using this information, the Pepper robot was able to avoid driving into obstacles even if the human signal appeared to instruct the robot to do so.

This is an interesting human-robot pairing that — while still in its early stages — I believe has applications in telepresence (the authors explain that this system could be used by sick or disabled individuals who can interact with family members or friends in a mobile way) or in remote navigation, as a more intuitive drive system for exploratory or recon robots in disaster zones or other inhospitable regions.