Marc Raibert of Boston Dynamics Interview – BigDog Robot Makers – By Sander Olson

Here is the Marc Raibert interview, attached. Dr. Raibert is a former MIT Professor who founded the Boston Dynamics corporation. Dr. Raibert and his colleagues have succeeded in developing hopping, self-balancing robots that can run and traverse terrain like animals. Boston Dynamics has recently unveiled a series of sophisticated mobile robots, including Big Dog:

Boston Dynamics has recently unveiled that it is the recipient of a DARPA contract to develop the Legged Squad Support System (LS3) robot, which will be an upgraded and improved version of Big Dog. LS3 will be able to walk 20 miles before needing refueling, will have a 400 pound payload capacity, and will use sophisticated AI to navigate its surroundings.

LS3 – Legged Squad Support Systems was funded by DARPA for $32 million

LS3 is a dynamic robot designed to go anywhere Soldiers and Marines go on foot. Each LS3 will carry up to 400 lbs of gear and enough fuel for missions covering 20 miles and lasting 24 hours. LS3 will not need a driver, because it will automatically follow a leader using computer vision or travel to designated locations using sensing and GPS. The development of LS3 will take 30 months, with first walk out scheduled for 2012. The development of LS3 is being funded by DARPA and the US Marine Corps

Question 1: Tell us about Boston Dynamics. How long has this company been developing robots?

Answer 1:We were founded in 1992, and the company started making robots 7 years ago, in 2003. Designing and making robots is now 3/4 of our business, with the other quarter consisting of modeling and simulation.

Question 2: Boston Dynamics recently unveiled the Legged Squad Support Systems (LS3) robot. Why is this robot special?

Answer 2: We had previously designed and built a robot called BigDog with funding from DARPA. BigDog was designed to explore the possibility of getting improved mobility with legs rather than wheels or tracks. That five year program was successful, and the LS3 robot is essentially BigDog on steroids. Both BigDog and LS3 have gasoline-powered engines. The LS3, when fully developed, will carry a greater load further than BigDog, will be more autonomous, and will have generally enhanced capabilities.

Question 3: How specifically does LS3 compare to BigDog?

Answer 3: Keep in mind that LS3 is just on the drawing board now and does not yet exist. But the goal is for LS3 to go way beyond BigDog’s capabilities. BigDog carries 340 pounds on flat terrain, and climbs difficult terrain carrying 100-125 pounds. It has a maximum range of about 12 miles, but not while carrying maximum payload. By contrast, the LS3 is designed to carry 400 pounds for 20 miles over rough terrain. The LS3 will weigh about 1000 pounds, including the weight of the robot, fuel and 400-pound payload. It will also be able to walk, trot, run, and wade through water. It will be able to operate at night, though probably with less mobility than in daylight. So LS3 will go far beyond what BigDog has done.

Question 4: Is this robot designed to be easily mass-produced? Will the military eventually be able to order these robots by the thousands?

Answer 4: The current LS3 contract is to build two prototypes. Additional work will be needed before a production version is available. But I see no reason why LS3 could not be produced in whatever quantity was required. In the long run I think the cost of an LS3 could be in the ballpark of a nice car.

Question 5: How many phases are there to this program? How much funding are you receiving?

Answer: Phase one will take 30 months to complete, so we should see LS3 models operating by the end of 2012. If there is a phase 2, it will begin shortly after completion of phase one, and it will probably run for eighteen months or two years. We will receive $32 million from DARPA for phase one.

Question 6: Will the LS3 robot have vision?

Answer: We are still in the process of designing LS3, but we anticipate it will have both a stereo vision system and a laser range finder system. These systems will allow it to sense the terrain around it, as well as sense the locations of people. A primary mode of LS3’s operation will be to follow a human leader. BigDog can follow a person using a laser ranging sensor under limited conditions. LS3 will be much more capable in this area, including the ability to do night operations, albeit with reduced performance.

Question 7: So the LS3 will be capable of semi-autonomous operation? What sort of GPS capabilities will it have?

Answer 7: LS3 will have a GPS system, but GPS works poorly in forests, jungles, and in cities with tall buildings where the GPS signals are obstructed.

Question 8: How will LS3 interact with people? Will it be able to respond to verbal commands?

Answer 8: The overall goal is to make it so LS3 requires as little attention as possible to operate. We can’t count on having a person dedicated to driving LS3 all the time, as is the practice for most deployed robots today. So we expect it to obey simple commands, such as “follow me”, “start following him”, “go there”, “wait here”, etc. The commands might be delivered through hand signals, verbal commands, a laser pointer, or through a hand-held device similar to a cell phone.

Question 9: Could the LS3 have any civilian uses?

Answer: A variant of the LS3 could be used in agriculture, in forest fighting, in emergency rescue, or possibly as an advanced sentry.

Question 10: Has Boston Dynamics done any work on humanoid robots?

Answer: Yes. From 1999 through 2004 we helped Sony Corp with their small humanoid robot called QRIO. And recently we have been building the PETMAN robot for the US Army. Unlike many humanoid robots that are designed for entertainment, PETMAN is designed to do specialized testing of clothing, so it has to have the shape and motion of the person.

Question 11: What sort of sensing and perception capabilities will LS3 have? Will it be able to tell the difference between a tree and a soldier?

Answer: Yes, it will be able to distinguish people from other objects, and to measure the shape of the nearby terrain. We have the NREC group at Carnegie Mellon and the Jet Propulsion Laboratory on our LS3 development team, and both of those organizations have extensive track records developing robot perception systems. They are both subcontractors to Boston Dynamics with responsibility for developing the perception system for LS3.

Question 12: Will LS3 have any tactile sensors, or hearing?

Answer: LS3 will have sensors in the legs that allow it to feel the terrain. That approach is used extensively in BigDog, and we will continue to use it for LS3. It might be possible to add hearing, but we don’t have a plan for that. LS3 is a mobile platform that can be upgraded in future versions by adding various payload modules, just as modern aircraft have upgradeable avionics modules.

Question 13: What other robots does Boston Dynamics have?

Answer: At any one time we work on about 10 different robots. There is LittleDog, a small four-legged robot used by leading universities for research in learning, There is PETMAN, a robot that walks like a man on two legs and will be used to test specialized clothing. RHex is a small six-legged robot that has great mobility in rough terrain, and it can also swim. We also have a climbing robot called RiSE, which can climb trees, walls, and fences, and we are developing a robot called SquishBot, that maneuvers through tight spaces. And there are a few others in the works you might hear about later.

Question 14: Is a lack of standards hampering robotics development?

Answer: I don’t think so. I believe that a comprehensive set of standards might even be counterproductive at this early stage of development. The technology is changing so rapidly that a series of standards might actually lock-in the wrong technologies. I think standards can be a good thing – they have certainly been good for the PC industry – but the kind of robotics we currently do are so free-form, I think they would just get in the way.

Question 15: Is Boston Dynamics doing any AI research?

Answer: We have a group that is developing some very powerful AI and we are applying it to our products. On the robot side, we use AI to make machines more agile, dexterous and easier to use. But we also embed AI in our human simulation products, which makes them useful for a far wider set of training applications.

Question 16: How do you see the robotics field developing during the next decade?

Answer: In the 10 years since 9/11, thousands of ground robots have made their way to Iraq and Afghanistan to help defeat IEDs. And the use of UAVs has grown extensively in that period as well. I think the success of these unmanned military systems has opened a door that will lead to a broader set of applications and opportunities in the next 10 years.