We've been tracking the Qbo open-source robot for over a year now, as creator TheCorpora aims to turn it into the Ford Model T of experimental robotics. Now there's a new video of the Qbo in action, with the endearing 'bot reacting autonomously to users by virtue of its computer vision algorithms, rather than using simple proximity sensors like, say, a robot vacuum cleaner.
The demo was designed to show the company eventually responsible for mass-production of Qbo exactly what the robot was capable of, though it actually ended up teaching TheCorpora more about naturalistic interactions. The log file storing the 'bot's activity recorded various occasions where bright sunlight distracted its vision, as well as a tendency to position itself further back from taller people than from shorter.
Rather than making a bad impression, however, the glitchy Qbo was perceived as more human in its behaviors. Just as people can be dazzled by bright light, or generally increase inter-personal distance when near taller companions, those tendencies in the robot only made it seem more alive rather than a PC on wheels.
The question TheCorpora is left with is deciding just how many bugs in their "social robot" software to iron out, and how many to leave based on the premise that "flaws" in perfect robot behavior actually contribute more than they confuse. That's only likely to make the software environment more complex; for this particular demo, only the artificial vision, synthesis and speech recognition, the chatterbot and the modules in charge of sending and receiving data to the two Qbo internal controllers were active, but there's considerably more in the roadmap.