The scene at the Shaw Conference Center in Edmonton, Canada, on July 31 was right out of the pages of The Pied Piper of Hamelin. But in this case, the piper was a pipe -- a big upright tubular robot on wheels -- and following its soundless allure was a throng of some 200 adult researchers. The scientists were eager to witness a milestone in robotics and artificial intelligence, hoping that a robot named Grace could carry out "her" ambitious assignment.
Without human guidance or a preprogrammed map, Grace was to find her way to the registration counter of the American Association for Artificial Intelligence's 2002 conference, sign in, navigate to the elevators leading to the conference rooms, "schmooze" with people in the elevator and hallways, take her place behind the podium, and deliver a lecture about herself.
Even on the eve of the AAAI event, Grace's chief creator, Reid Simmons of Carnegie Mellon University, was hesitant about the robot's upcoming performance. He said he'd be happy if Grace accomplished even half of its mission. "If we just had another month or two to prepare, I'd feel more confident," he said before leaving Pittsburgh for Edmonton.
10-YEAR CHALLENGE. Simmons needn't have worried. Grace came through with flying colors, although she took five minutes longer than he expected -- a total of 50 minutes from when she was dropped off outside the conference center to the time she reached her speaker's podium.
"Everything was still a little ragged," admits Simmons, a senior researcher at Carnegie Mellon's Robotics Institute. "The most frustrating thing was her speech-recognition program," especially when she tried to engage in small talk during the schmoozing procedure.
The tasks that Grace carried out autonomously were first laid down in the 1999 Mobile Robot Challenge, issued by the AAAI to raise the bar for intelligent robot behavior by creating "social robots." In what was envisioned as a 10-year effort, such machines would be capable of interacting with people on human terms, not only engaging in conversations but also knowing how to conform to social etiquette, like standing patiently in a line at a registration desk.
NOT READY FOR Q&As. One other robot, from Boston's iRobot Corp., completed the entire routine, but its actions were remotely controlled by a human operator. Its trek from the building's entrance to the podium took 20 minutes compared to Grace's 50. "That's now our next goal -- to do it as fast as a human-operated robot," says Simmons.
But one aspect of the robot challenge may take a few more years to achieve: responding to technical questions from the audience at the end of the two-minute lecture.
After the AAAI issued its robot challenge, recalls Simmons, "We decided to take a stab at it. But we knew we couldn't do it alone," so he enlisted four other outfits to help. Grace's historic performance is the result of an on-going collaboration among the Naval Research Lab (NRL), Metrica Inc. in San Antonio, Swarthmore College, and Northwestern University.
The NRL is in charge of speech recognition. Metrica handles vision-based chores such as recognizing gestures and tracking people. Swarthmore heads the work on reading signs and name tags so Grace could find her way around the conference center. And Northwestern enabled Grace to give the talk about herself. "It's very unusual to cross institutional boundaries in robotics research," Simmons notes, "but the results speak for themselves."
JUST LIKE BLANCHE. For now, one thing Grace lacks is an arm. Simmons hasn't yet enlisted any mechanical engineering experts, and the group decided that as AI researchers, it would be best to dispense with appendages. When it comes to opening doors or pushing elevator buttons, Grace, like Blanche DuBois in A Streetcar Named Desire, depends on the kindness of strangers.
"Could you pin the badge to me?" Grace asked the person behind the registration desk. "I'm afraid I'm all out of hands." By Otis Port in New York