How machines like the LS3will change theatres of war
“The weight of your equipment is proportional to the time you have been carrying it.”-Canonical Murphy's Laws of Combat
Abstract:
Boston Dynamics has been a source of game changing robot technology over the past decade. The company created the BigDog program, a robot that has demonstrated animal-like locomotion. It has created an enormous amount of appeal from military customers and academic institutions alike, both of which have accelerated its development for future field operations. The newest iteration of the program named the Legged Squad Support System (LS3) is expected to be delivered in 2015.
Introduction:
Every day soldiers out in the field bear an enormous burdenfrom the equipment they have to carry. The military has used and are still using tank-like treaded robots, most of which are not autonomous and require a human operator. For a vast majority of infantry operations there is a fundamental desire for vehicles that can lessen the physical amount of cargo soldiers must carry. This is not limited to operations occurring in rural areas, because as Professor Stefan Schaal of USC’s Computational Learning and Motor Control Lab commented, “Rough terrain can be found in many daily life scenarios … a disaster site, a kids playground, or even a toy-cluttered room in a daycare center is currently not suited for existing robots,” making the mission need for legged robots that much more important [1]. Boston Dynamics, with funding from the Defense Advanced Research Projects Agency (DARPA) and support from other research universities, made a concerted effort to lessen the physical hardship the modern warfighter can encounter whether it be from carrying equipment or evacuating wounded soldiers. In 2005 they created the first of what would be a line of four-legged robots, the first of which they called BigDog. While BigDog was considered a success, the creators expressed a desire for better “planning” strategies and a design that could carry a heavier payload and allow the robot to self correct itself should it fall over while in the field [2]. The requirement for better “planning” algorithms spawned the Learning Locomotion project, and both requirements manifested themselves in the successor generation called Legged Squad Support System (LS3). Both programs aimed to further advance the field of robotics while delivering sophisticated robots and planning procedures to meet the needs of their military customers.
BigDog
The quadruped, shown in Figure 1,weighed 240 lbs., and measured 1 meter tall, 1.1 meters long, and 0.3 m wide.The robot resembled a small horse, but it featured front legs that flexed similar to a dog’s. Additionally the robot hosted an array of fifty sensors used for maintaining balance and for successfully maneuvering through uneven terrain. It contained an onboard computer that would digest high level operator commands that would direct the robot. This allowed an operator to direct the robot to turn left ninety degrees for instance by simply turning the joystick of the “operator interface” controller and the robot would make all the minute adjustments of its legs to accomplish that command. The robot was demonstrated walking across multiple types of terrain including muddy forest, snow covered hills, and slippery ice. As a substitute for truly autonomous operation, BigDog also makes use of a laser radar or LIDAR. The difference between traditional radar and LIDAR is that traditional radar emits radio pulses and LIDAR instead uses a laser which allows for better precisionand for BigDog translates into better knowledge of its surroundings. In this configuration,illustrated in Figure 2, the leader would be wearing reflective strips on the posterior of their backpack equipment and based on the results of the LIDAR BigDog would attempt to follow at a fixed distance.
Figure 2: LIDAR “follow the leader” configuration [3]
USC Connect
The University of Southern California is among the six research universities (Carnegie Mellon University, Florida Institute for Human and Machine Cognition, Massachusetts Institute of Technology, Stanford, and University of Pennsylvania) selected to receive DARPA funding to investigate legged locomotion through the Learning Locomotion project. Boston Dynamics provided each of the universities a LittleDog shown in Figure 3, and released an application program interface (API) to allow the different universities to modify the control software of the robot.The physical dimensions of the robot were scaled down from its BigDog relative, measuring only 34 cm in length and a standing shoulder height of 12 cm [4]. The intent of the program was to give the miniaturized robot a way of learning about terrain it would encounter and over time make better decisions for foot placement and increase its probability of crossing terrain successfully[1]. USC and the other funded universities competed in tests developed by DARPA in order to measure the success of the different universities’ techniques, one such test is shown in Figure 4.
Figure 4: LittleDog crossing a rock obstacle course [6]
In 2009 USC competed in DARPA’s final test scenarios and was one of the four finalists to complete the final tests. The USC team outperformed “final government program guidelines” due to their “machine learning” software components and the components that allowed for the robot to perform jumps and slides [1]. The team accomplished this by using a technique called “supervised learning (SL)” which in contrast to “reinforcement learning (RL)”, observes an expert demonstration after which the team’s algorithm would select footholds with similar characteristics. In RL, the robot would learn about footholds based on its own successes and failures passing a terrain.Overall when the terrain is more difficult, the SL approach tends to perform more reliably because the RL approach will not be able to learn from its failures, and the SL approach allows for more direct feedback [5].Techniques like the aforementioned were developed by the member universities and each contribution made an impact on the software improvements of the next BigDog generation.
LS3: The Next Generation
The next generation of BigDognamed LS3 shown in Figure 5,was made to improve on its predecessor in terms of how much weight it could carry and toincorporate enhancementsthat would prepare the robot for military usage. It is expected to be able to travel 20 miles without refueling, have a carrying capacity up to 400 lbs., and will carry auxiliary power sources for the purpose of charging radios and other mission electronics [4]. Additionally they modified the engine the robot would use in order to partially qualm the noise emissions during operation. To complement the physical alterations, the LS3 has been given software upgrades in the form of three new autonomous settings: Leader-follower tight, Leader-follower corridor, and Go-to-waypoint where the robot is provided GPS coordinates to traverse to. Voice commands to initiate the previous autonomous settings have also been incorporated [4].DARPA began working on this phase of the project in September 2009. The total project cost is estimated to be $54 million and is expected to be put into operations in 2015 [7].
Challenges
Despite the advances that have been made in the LS3 generation, work is still being done to refine the robots ability to closely “follow the leader” and if necessary because of its larger size, “follow its own chosen path that’s best for itself” in order to eliminate the need for a human operator [7].
Additionally the increased carrying capacity for the robot presents difficulty for the robot’s leg structures if the robot were to need to recover from tipping over while traversing hilly terrain. As weight requirements are increased the team that manages structures must determine if the robot can tolerate the increased load without compromising its operational integrity. These integrity checks include all locomotive elements from the motor, to the material used to construct the robots legs, to the springs that are used for suspension.
Lastly the robot itself remains susceptible to firearms and while it is uncertain that it would directly be utilized in planned combat operations, the threat of unexpected attacks still exists. As of September 2013, DARPA offered $10 million in additional funding to create a bulletproof version of the LS3, and while there is less concern about the cargo portion of the robot, the “head” portion that facilitates navigation is of primary concern [2].
Conclusion
These three iterations of the DARPA’s Legged Locomotion project demonstrate the complexity of accomplishing a task that we as humans take for granted every day. They also serve as examples of how animals in nature can inspire some of the most technologically advanced machines in existence today.Once refined, the LS3 will present warfighters with a welcomed advantage, one that willfundamentally save lives and ease their day-to-day physical burden.To complement the military advancements that culminated into the LS3, the academic robotics community has also benefitted from this project’s existence, and I would venture that the relationship between industry and academia will continue to strengthen as the capabilities of these robots continues to evolve.
References
[1] "Teaching a RoboDog New Tricks." USC.Viterbi School of Engineering, 23 Oct. 2009.Web. 14 Nov. 2013.
[2] Ackerman, Evan. "Boston Dynamics Gets $10 Million from DARPA for New Stealthy, Bulletproof LS3." IEEE Spectrum. IEEE, 23 Sept. 2013. Web. 16 Nov. 2013.
[3] Raibert, Marc, Kevin Blankespoor, Gabriel Nelson, and Rob Playter. BigDog, the Rough-Terrain Quadruped Robot. Tech. Waltham: Boston Dynamics, 2008.
[4] "Legged Squad Support System (LS3)." DARPA RSS.N.p., n.d. Web. 15 Nov. 2013.
[5] Kalakrishnan, Mrinal, Jonas Buchli, Peter Pastor, Michael Mistry, and Stefan Schaal. "Learning, Planning, and Control for Quadruped Locomotion over Challenging Terrain." The International Journal of Robotics Research 236th ser. 2011.30 (2010): n. pag. Sage Journals. Sage Publications, 12 Nov. 2010. Web. 15 Nov. 2013.
[6] Murphy, Michael P., Aaron Saunders, Cassie Moreira, Alfred A. Rizzi, and Marc Raibert. "The LittleDog Robot."The International Journal of Robotics Research 145th ser. 2011.40 (2010): n. pag. 7 Dec. 2010. Web. 14 Nov. 2014.
[7]Cronk, Terri M. Robot to Serve as Future Military's 'Pack Mule' Rep. Federal Information & News Dispatch, Inc., 19 Dec. 2012. Web.
[8] "The Day the Marines Met Their Robotic Mule." Popular Mechanics.N.p., n.d. Web. 05 Dec. 2013.