Colin Angle | Tuning in to robot wisdom
Interview withColin Angle, CEO of iRobot Corp.
- By Joab Jackson
- Jun 21, 2006
Colin Angle, iRobot Corp. CEO
Chatty, multitalented, walking gizmos such as Ronnie the Robot or R2-D2 come to mind when people think of robots. In fact, the leading robot builder today, iRobot Corp. of Burlington, Mass., found success shedding such anthropomorphistic perceptions. Since its start in 1990, the company has built software-driven, semi-autonomous machines'call them robots if you will'for specialized tasks. The company's handy-as-heck Roomba Vacuuming Robot has become a cultural icon. But iRobot also has delivered more than 300 tactical mobile robots, called PackBots, to the Defense Department for executing dangerous tasks.
CEO Colin Angle started the company while still in college, with two other robot enthusiasts, fellow Massachusetts Institute of Technology student Rodney Brooks, who is now chief technology officer, and Helen Greiner, now chairwoman of the board. Angle talked with GCN about where most robot researchers went wrong in their development and how the PackBot is used in Iraq.GCN: How did iRobot's approach differ from other robot companies at the time?
Angle: Early in the robot industry, there was this idea that a robot had to follow this serial progression, called Sense, Plan and Act. Sense refers to using the robot's sensors to create an internal three-dimensional reconstruction of the environment. Then, based on that internal model, the robot should figure out what exactly should happen and create a precise plan to move.
By following this strategy, researchers were creating these unwieldy and wildly expensive systems. The problem is that it is nearly impossible to create this sort of Cartesian internal model. The internal model rapidly becomes obsolete [because] we live in a dynamic world. Creating a detailed plan is also very difficult because it requires the robot to be very precise. Both of those are difficult assumptions to actually follow through on.
So Rodney Brooks looked at that and said, 'Let's think about this thing in a different way.' Insects survive pretty well, and they don't require some vast amount of computational power. So Rod developed a notion of using the environment as its own model. Rather than creating a plan based on an internal model, let's see how far we can get with heuristics and strategies. Only when we hit a wall, if you will, do we try to figure out what we need to model.
Certainly, autonomous tasks like vacuuming a room require lots of strategies and heuristics. Much of Roomba's software [is devoted to] figuring out what to do when a Roomba gets stuck, and how it should get itself unstuck.
Before Roomba, if people were asked what a robot vacuum cleaner would look like, I don't think anyone would have answered correctly [see photo].GCN: How did PackBot come about?
Angle: PackBot came out of the Tactical Mobile Robotics program at DARPA. Going into urban environments is one of the most dangerous things a soldier is asked to do. Can we create a robot that can take on some of this task?
DARPA invited us to write a proposal on how we would create such a robot. We didn't just write a proposal, we built a prototype. Our competition showed up with plans for very, very complex devices. We actually brought a robot, so if our reviewer said, 'That can't work,' then we could say 'Actually, it can. Here, watch.'GCN: How is PackBot used in Iraq?
Angle: We have over 300 robots in Iraq today. They have performed well over 100,000 missions. The dominant use is in explosive ordnance disposal. When a suspected improvised explosive device was found, the bomb squad was called and an armored soldier went up to the device to evaluate its state.
Not an optimal strategy, from the safety of the soldier's perspective. So Defense Department teams have aggressively adopted the robot as the preferred method of addressing [IEDs]. PackBot drives up to the suspected device and sends back video and audio information to the operator, who is managing from a hardened laptop.
It has a manipulator on the front so it can pull wires. It can excavate, dig around and figure what is going on. Then it is capable of rendering that device safe in one of a few different ways. If there are wires, it can pull those out. If the IED is well-understood, it could fire a flood of water through the bomb and destroy the detonation system.GCN: So the PackBot is extensible.
Angle: It is a platform. You can bolt on other systems, connect them to the serial and parallel ports, and to the video ports, and so forth. PackBots have also been in Afghanistan, clearing caves. They were used to go enter buildings in advance of the troops and feed back situational-awareness information. These PackBots had a basically simple and rugged audio/video head, which was bolted on. But you could take off the head and put a much more sophisticated payload on it, such as a seven-foot manipulator that can carry four cameras, or fiber-optic spooler that can be bolted on the back, in case you want to have a direct connection to the robot rather than using wireless transmission.
At one point, there was a need in Iraq to bolt on chemical sensors. They were finding mass grave sites there, and the [former Iraqi government's] strategy to discourage evacuation of grave sites was to pour toxic chemicals and other bad stuff over them. So the robot characterized what was waiting for soldiers without putting them at risk.
These are all different payloads that have been developed. We actually have a developer conference for the PackBot. [Last November], it was attended by over 100 potential developers.GCN: Did you pick up any lessons from the most recent DARPA Grand Challenge? [Last fall, teams built autonomous vehicles that successfully completed a 131-mile road race.
Angle: We sent down a significant team to watch. It was a great recruiting event for us. There were certain restrictions on technology and funding, so we opted not to enter a vehicle. But the students who created [these systems] really got a fantastic education in creating practical robot devices.GCN: So DARPA director Anthony Tether was correct in assuming the Grand Challenge would inspire youngsters to start thinking more about robotics?
Angle: You couldn't ask for a better place to start looking for people.
Joab Jackson is the senior technology editor for Government Computer News.