DARPA's Robotic Mule Could Be in Troops' Hands by 2014

The Pentagon’s experimental robotic pack mule underwent two weeks of field testing in December at Ft. Pickett, Va., the first in a series of trials that could have it hauling Marines’ gear within two years.
The Legged Squad Support System, or LS3, represents the latest evolution of a technology the Defense Advanced Research Projects Agency has poured $54 million into since 2009.
“The purpose of the LS3 program, first and foremost, is about solving a real military problem: The incredible load of equipment that soldiers and Marines have to carry in Afghanistan,” said Army Lt. Col. Joseph Hitt, DARPA's LS3 program manager. “The purpose is to offload some of that onto a robot that behaves like a mule, like a trained animal.”
DARPA’s goal was to build a robot that mimics the load-bearing abilities of an actual mule, so troops could pile their gear onto it while traveling. More specifically, they aimed to create a robotic platform that could carry multiple 40-pound packs and walk at least 29 miles in a day without intervention, Hitt said. The mule eventually will provide auxiliary power to troops on the move, reducing their reliance on bulky batteries.
Dismounted troops routinely carry up to 50 pounds in protective gear alone. Adding in ammunition, food, communications systems and batteries, they are often saddled with 100 pounds or more.
“That can cause soft-tissue injuries,” Hitt said. “We’re getting back to the original purpose of the program, which is to offload the burden from soldiers.”
The LS3’s recent two-week exercise is the first time the latest iteration of the system has gone afield. About the size of a large dog,the LS3 walks around on four piston-driven legs that give it a sort of jolting gait. It can also jog up to six miles per hour over smooth terrain.
“The exercise was highly successful because it showed us not only what the robot can do, but, most importantly, what it can’t do,” Hitt said.
Weakness of perception and the robustness of the design were two issues singled out during the trials. The perception problem was not rooted in its sensor, but rather its “understanding” of the data collected about the surrounding environment, Hitt said.
“The sensors we have are fine,” he said. “We need to bring in the sensor input … then the robot has to be able to generate a path through that environment and determine how it should behave while navigating through it." As the mule is introduced to new obstacles and varying terrain, situational awareness “is a continual challenge for us,” he added.
Engineers also must toughen up the robot before it deploys with Marines.
“We need to make sure the platform is robust enough to take the abuse that Marines are sure to give it,” he said.
So far, LS3 has performed successfully in both wooded and urban environments, the two scenarios tested at Ft. Pickett.
The LS3 demonstrated several firsts during the latest trials. The robot was able to move to a predetermined GPS coordinate, navigating around obstacles on its own, in total darkness. That capability will allow the system to travel with and resupply units during night operations, Hitt said.
Using a laser range-finder, specialized cameras and stereo vision, the robot tracks its human leader and is able to generate a “vision” of surrounding terrain. A video released Dec. 18 shows the LS3 negotiating a meandering creek bed, overstepping logs and negotiating moderately dense forest. At one point it falls trying to cross a small ravine, does a complete barrel roll, rights itself and continues on.
Engineers also debuted a voice-operation system that allows a human “leader” to operate the robot with preset commands like “follow,” “stop” or “sit.” The mule currently recognizes 10 relatively broad commands.
“Our philosophy is that rather than developing an infinite number of specific commands, we wanted to provide a few very reliable and robust commands and have the user adapt those to different scenarios,” Hitt said. The system was kept relatively simple for ease of use. Because the voice commands are similar to those used to train and handle domestic animals, it was “surprisingly easy” to integrate the voice-operation system, Hitt said.
Troops also prefer to communicate with verbal commands, so integrating a robot that will respond to a soldier’s voice makes sense, he added.
Given the command to follow its leader, LS3 can either tag along on the same path or be told to find the most efficient route it can find. The robot can also be told to go to a GPS waypoint and will navigate itself there without human assistance.
The latest iteration of the LS3 is also much quieter than preceding models. Its noise signature is now a “tactically relevant” 70 decibels — between the volume of a normal conversation at three feet and a telephone dial tone.
Hitt said the system’s capabilities are the legacy of a decade of DARPA research into robotic autonomy and perception in programs like the Big Dog and Crusher robots. Breakthroughs made with LS3 are feeding directly into other programs, like the super-fast Cheetah and Atlas Humanoid  robots, he said.  
“LS3 would not be where it is today without those programs,” he said.
Plans are to repeat similar field trials every three months for the next two years, after which the robot will be handed over to a Marine unit for real-world use. The robot will be introduced to new environments and new challenges at each location.
Next up is a two-week test in March at Marine Corps Air-Ground Combat Center 29 Palms, Calif. Other tests are scheduled at the Marine Corps Mountain Warfare Training Center in Bridgeport, Calif.; Camp Lejeune, N.C. or Camp Pendleton, Calif. and at Ft. Benning, Ga.
Final tests will be conducted at a training exercise to be selected by officials at the Marine Corps Warfighting Laboratory. That test is tentatively set to coincide with the 2014 Rim of the Pacific exercises.
“Then we will be able to hand LS3 over to a Marine company and for us, the engineers, to step aside and say, ‘Here you go,'” Hitt said.
 
Photo Credit: DARPA

Topics: Robotics, Unmanned Ground Vehicles, Science and Engineering Technology, DARPA