The NAO robot that we’ve seen in the past (having received a sexier body, in addition to playing nice with Kinect) is back in the limelight – thanks to researchers at the Albert-Ludwigs-Universität Freiburg’s Humanoid Robots Lab. The culmination of their work was presented at ICRA 2011, where it ought to be of interest to NAO developers. This particular lab worked on the laser range finder attachment with Aldebaran Robotics, where it has ended up as a method for detecting obstacles by merging data from that attachment with NAO’s monocular vision. Since the laser range finder is positioned on top of NAO’s head, it isn’t possible to detect obstacles right in front of NAO (a case of robotic far-sightedness, perhaps?), hence requiring the robot had to make frequent stops to scan ahead. As for the new method, it will rely on camera images as much as possible so that time-consuming laser scans can be avoided.
The laser will scan the ground plane first, and anything that extends above it will be considered to be an obstacle that will be classified based on their color and texture in order for NAO to detect them using the camera alone. NAO is able to follow that up by building an occupancy grid to plan its path.
What next for NAO?