A handy collaborator
THE most popular event at the International Conference on Robotics and Automation, held in Seattle at the end of May, was the Amazon Picking Challenge. Dozens of humanoid robots competed to move items ranging from rubber ducks to paperback books between shelves and a plastic bin, thus simulating the process of assembling an order at the retailer’s warehouse. An enthusiastic crowd of academics and roboticists gathered to applaud every success. In the event their cheers were few and far between. Many robots failed to grasp a single item. Even the winner picked just ten during its 20-minute test.
The problem is that although robots are good at precise, complex activities like welding a car, they are terrible at tasks humans find trivial, such as recognising objects and planning how to navigate or work around them. Building a robot that can move (slowly) through a home or workplace also requires lots of computing power, sophisticated actuators (a type of motor), a host of sensors and a hefty battery. Little surprise, then, that some of the robots in the Amazon competition cost as much as half a million dollars.
A more sensible solution, according to Walterio Mayol-Cuevas of the University of Bristol, in England, is to use people for navigation and planning and give robots the freedom to do what they do best. In a paper he presented at the conference, Dr Mayol-Cuevas described an intelligent hand-held robot which he and his graduate student Austin Gregg-Smith have built. Using a grass strimmer as a base, they designed a tentacle-like robotic arm that can move freely in any direction. Motors pull cables that flex a carbon-fibre rod, at the end of which a variety of tools can be attached. In the prototype, these included a magnetic gripper for placing floor tiles in a specific arrangement and a virtual paintbrush for colouring on a computer screen. A human operator carries the robot to the right place and pulls its trigger. The robot then goes to work automatically, twisting its head to paint a preprogrammed picture or to grab a tile of the colour needed to complete a pattern.
Dr Mayol-Cuevas and Mr Gregg-Smith have also built a second prototype, with a spinning brush at its tip. This, they suggest, could make the perfect cleaning tool. As Dr Mayol-Cuevas observes, human cleaners can get sloppy—not scrubbing thoroughly enough or, perhaps, missing a spot. In a place like a hospital that can matter a lot. But a robot can keep track of what has and has not been cleaned and direct the operator there. It may even be able to reach into spaces that are inaccessible to a person.
Dr Mayol-Cuevas’s robots employ basic gestures to communicate with their users, such as pointing at places where they would like to be put to work or shying away from tiles of the wrong colour. Tests suggest the devices reduce both the time required to complete a task and the difficulty of doing so. Dr Mayol-Cuevas thinks that, as well as helping experienced operators to do a better job, hand-held robots will eventually let unskilled users tackle difficult tasks like bricklaying. He hopes that might lead to civic “crowd-building” efforts, in which passers-by simply pick up smart tools and contribute to the construction of new houses and schools.
At the moment, though the prototypes have internal batteries, they rely on external computers and motion-capture sensors to provide the information needed to guide their human operators. However, the two researchers hope to change that for their next robot by incorporating a Kinect, a motion-tracking device developed by Microsoft for use in video gaming. Dr Mayol-Cuevas also wants to improve the robots’ communication skills, perhaps by adding a small screen or giving them the power of speech. Whether that is wise remains to be seen. The idea of having an argument with a paint brush about whether a wall has been painted properly sounds surreal.