BigDog Steps Into Action
Quadrupedal packbot pushes boundaries of locomotion technology to tread where other machines cannot.
The goal of the U.S. Army’s BigDog program is to develop four-legged robots capable of following soldiers across any kind of terrain. Roughly the size of a large dog or small mule, the robot is intended to carry several hundred pounds of supplies and equipment to lighten troops’ loads in combat.
Dismounted infantry may one day rely on four-legged robots to carry equipment and ammunition into battle. The U.S. Defense Department envisions the machines following troops into rugged terrain or through densely packed urban areas too confined for conventional vehicles. These automated quadrupeds are part of a larger government initiative to study how animals move and to apply those characteristics to robotic systems.
The BigDog program seeks to develop robots that can carry several hundred pounds of equipment, follow soldiers at set distances and respond to a variety of commands. Administered by the U.S. Army’s Tank-Automotive Research, Development and Engineering Center (TARDEC), in
The program is developing quadrupedal systems that can run and scramble as opposed to carefully placing steps in a structured gait. BigDog is meant to move quickly over steep hills, rubble and muddy ground. “It [the robot] really doesn’t know what it’s stepping into until it places a foot there. That is the challenge of quadrupedal motion in a nonstructured environment,” he says.
Last year, TARDEC launched an offshoot BigDog program by starting a small business innovation research (SBIR) program with Boston Dynamics, a Cambridge, Massachusetts-based robotics and human simulation firm. Known as the BigDog Mule System, the project shifts the legged locomotion research from DARPA to Boston Dynamics. Andrusz explains that the goal of the mule system program is to load the robot with several hundred pounds of supplies and have it follow a dismounted soldier over rough terrain.
The SBIR program has progressed, but future milestones include increasing the weight that can be carried and demonstrating the ability to follow a soldier. The project is raising the difficulty level of these steps incrementally over time, Andrusz adds. For example, the initial goal is for the robot to follow an individual at a preset distance down the same path. Researchers then want to increase or decrease this distance with a command. “If a soldier wants the robot to follow at a certain distance, he can say ‘follow at 100 meters or 50 meters,’ or ‘move closer because I need some of the stuff on your pack,’” Andrusz shares.
After the soldier-following capabilities are fully developed, work will focus on methods to enable the robot to deviate from the soldier’s path. Warfighters moving through extremely difficult or dangerous areas can signal the BigDog to examine its surroundings, deviate from its planned path and take an alternate route to reach an objective.
Researchers also are trying to increase the robot’s speed and distance traveled on rough terrain. “Our goal is to have a robot that can do anything from a slow walk to follow a soldier to sprint as fast as it can go across a certain area. That may be to draw fire or to evade fire and give aid,” Andrusz says.
The latest BigDog robot features two sets of legs with knees that bend in opposite directions. Andrusz notes that the original version had the knee-joints bending in the same direction, similar to an animal, but additional research demonstrated that mobility improves if the legs’ motion is disparate.
A gasoline engine powers a hydraulic actuation system that moves the robot’s legs, which also feature shock-absorbing elements designed to recycle energy from one step to the next. Additional work is taking place for a new engine for the robot because the prototype is powered by a model airplane engine, which is inefficient and loud, Andrusz explains.
|Because it is designed to operate on difficult and unpredictable terrain such as rubble and mud, the BigDog robot relies on gyroscopes and sophisticated software to maintain balance. A process known as dynamic balancing allows the robot to immediately compensate should it loose its footing.|
BigDog relies on a technology called dynamic balancing, which allows the robot to immediately compensate if it loses its footing or if it is pushed or shoved from the side. Andrusz explains that dynamic balancing is vital for any kind of legged locomotion, but it is especially crucial for operating in difficult terrain. He adds that Boston Dynamics engineers have tested the prototype on muddy ground, where it maintained its balance when shifting from firm terrain to puddles up to three inches deep. Currently, if BigDog falls over, it cannot get back on its feet. But he says that it can recline into a parked position from which it can rise.
Writing software that allows the robot to maintain its balance was another challenge because the machine has to be told what to expect. “You can’t tell it to do something about slipping. It already has to know. There has to be an algorithm in place that compensates automatically. Tell it after the fact, and the system will probably fall down,” he emphasizes.
The robot’s sensors fall into two groups: mobility and soldier following. Andrusz indicates that the DARPA portion of the program is developing the mobility sensor suite. For soldier following, researchers are examining stereovision and other options such as laser deletion and ranging (LADAR) and scanning systems that record where soldiers have placed their feet. He notes that a proof-of-concept vision system may rely on an obvious marker such as brightly colored shoes for the machine to detect. But as the system progresses, the technology will evolve to follow a person’s footsteps without aid.
In this area, researchers are trying to stay away from solutions that involve approaches such as radio frequency (RF) detection to avoid battlefield emissions issues. “You don’t want humans to emit anything on a battlefield—that’s dangerous. The other problem is that the next soldier the robot might follow might not have an RF tag, or the tag might get lost,” he explains.
The program also seeks to advance soldier-recognition technologies. Andrusz says that this can be achieved with stereo imagery to judge distance between soldiers and LADAR distance-sensing systems. The current emphasis is on a combination of LADAR and stereovision.
At the end of the two-year SBIR initiative, TARDEC will decide whether to proceed with more sophisticated developments or to step back and re-examine the technology. The program’s main deliverable will be a demonstration of the robot’s capabilities. Andrusz explains that this is an ongoing research and development program and that Boston Dynamics will continue working on other four-legged systems.
The demonstration is tentatively scheduled at a TARDEC facility in
If the locomotive technology is successful, it will allow the robot to follow soldiers across terrain that wheeled or tracked vehicles cannot traverse, including highly unstructured areas. Andrusz states that TARDEC’s Robotics Mobility Laboratory specializes in various locomotion systems. “Our lab specifically concentrates on smaller vehicles—500 pounds and under. It’s very hard to increase mobility at that size. When we look at quadrupeds like BigDog, what we’re looking at is ultrahigh mobility. After we get it to follow a path outlined by a soldier, can it now go through the woods and step over logs and through underbrush without tripping itself up? Can it go into an urban environment with very complex terrain? Let’s say there’s been an earthquake or it’s a war zone and there’s building rubble—can it traverse those types of situations?” he offers.
Other firms such as Yobotics Incorporated in
A major difference between BigDog and other machines such as Sony Corporation’s Asimo bipedal robot is that from its inception, BigDog has concentrated on functioning on terrain that does not allow the robot to know where its next step will land. Boston Dynamics researchers were not trying to develop a system that could work in a controlled environment. “Scientists were experimenting on a tile floor but always with the assumption that the foot does not know what it’s going to touch until it touches it. And it doesn’t know if it’s going to hold or slip if it does so. Dynamic foot placement was always a part of their work,” he explains.