For the last 25 years Prof. Veloso has been trying to create autonomous intelligent robots. Her IROS 2016 plenary talk on Tuesday shed light on her experiences in developing these intelligent robots, namely the mobile service robots – cobots – that have been navigating all nine floors of her building.
Her talk centered on some key questions regarding these autonomous machines, such as the problem of peer interaction with humans. That is, because these robots are not solitary figures on Mars or in a volcano or a nuclear reactor but are rather in a building with humans, it was critical to get this right. Getting the robots to be robust was the objective, but doing so wasn’t easy, especially because any environment with humans is very dynamic. The robots use the local features and perceive them as “short-term” versus “long-term” features. They have also developed methods for humans not to have to program the robots so that all they need to do is to instruct the robot using “quite defined language” and what she calls instruction graphs.
One fascinating aspect of the robots is something called “symbiotic autonomy.” To illustrate this, she refers to the cobots whose job is to escort visitors, among other duties, to her office. Although they are completely mobile, their functions are limited in other ways – they can’t push the elevator buttons or make a cup of coffee, for example. They do, however, proactively and autonomously ask (a human, the web, or other robots) for help with directions or for example, to push the elevator button. The robot accumulates the knowledge and learns from his interaction with humans.
So, what happens when there aren’t any humans around to help them When this happens, they light up and begin blinking in different colors, called “expressive lights.” People then approach the robot and ask the robot what it needs help with.
Prof. Veloso also spoke about the issue of transparency; that is, being able to “see” what the robots were doing – even without having to actually view them – by making use of semantic annotations and maps, as everything is digitized and logged.
The challenge is language, Prof. Veloso stresses – being able to translate robot language into human language – or something that humans can understand.
But it’s one that needs to be resolved, as “one day our lives will be surrounded by autonomy,” she says.
Manuela M. Veloso is the Herbert A. Simon University Professor in the School of Computer Science at Carnegie Mellon University. She is the Head of the Machine Learning Department and has joint appointment in the Computer Science Department.
IROS 2016 is being held at the Daejeon Convention Center (DCC) in Daejeon, Korea.
