Center Builds Robots With More Bang for the Buck
Organization focuses on building and testing prototypes quickly and including program sponsors earlier in the process.
The Crusher unmanned ground combat vehicle (UGCV) performs during a field test. When raised, the long mast enables surveillance capabilities.
As the demand for robotics expands in both the commercial and public sectors, developers at a university institute are working to move relevant technology into the marketplace rapidly. Engineers are creating smarter systems that are more autonomous and that have applications ranging from agriculture to combat. Current programs are spawning new ideas, and program officials are seeking to demonstrate technology to funding authorities quickly to determine the best path forward early in the development cycle.
The National Robotics Engineering Center (NREC), a unit of the Carnegie Mellon University Robotics Institute,
Compared to the remainder of the Carnegie Mellon campus, the center is staff heavy, employing 60 to 70 professional engineers and about a half dozen of both faculty members and graduate students. The mission of the engineers is to create functioning products, not to publish or develop new theory. The NREC also has on staff several high-level technicians whose expertise enables the center to integrate complex systems right at its facility. Officials at the NREC strive to create prototypes early to give program sponsors concrete ideas about how devices will operate, and they keep processes transparent, inviting government and company personnel to view their work.
Steve DiAntonio, director of business development at the NREC, explains that currently, few applied robotics capabilities exist in the marketplace in either the public or private sectors. Only in the past 10 years has there been a need to use laboratory-grade technology to make robotics more robust and cost effective so that the automation can be applied in true operational environments. By displaying capabilities in these types of environments, NREC personnel give interested parties a better idea of whether a technology should advance to the next stage and whether it would be productive. These same decisions are more challenging to make in the early technology stages. “It’s just difficult to connect the dots,” DiAntonio says.
The NREC prides itself on quickly developing concepts and then prototypes and delivering them in a real environment under realistic conditions. For example, in a coal-mining project, the NREC develops sensors for certain machinery. The center ensures sensors are fireproof and has them certified by the necessary regulating agencies so the devices can be tested in the real product environment.
In terms of U.S. Defense Department projects, the NREC has a technical readiness level of 5 to 6, meaning that it develops systems that have been tested in operationally relevant environments. One of the major military projects being conducted at the center is the Crusher unmanned ground combat vehicle (UGCV) for the UGCV PerceptOR Integration (UPI) program. PerceptOR stands for Perception for Off-Road Robotics. The UPI program combines Crusher with advanced perception, autonomy and learning techniques. The program incorporates system design across vehicles, sensors and software, enabling one component’s strengths to compensate for weaknesses in a different component. The UPI is a U.S. Army Future Combat Systems feed program.
The Crusher component requires intensive field experimentation. Approximately three times a year researchers take the systems to a challenging outdoor environment at locations such as
During the field experiments, the NREC invites personnel from Future Combat Systems, both military and contractor, to view the activities. DiAntonio shares that the NREC tries to give the Army some visibility into what Crusher is actually capable of doing so the service can make a judgment about whether to continue to invest in the technology. Allowing the military to decide future funding earlier in the development process aligns with acquisition officials’ calls for more prototyping before the Defense Department advances programs into system development and demonstration. The change in procedure saves funding and helps decision makers better determine whether a technology is robust enough to transition to a full program. The NREC has geared itself to excel at this type of advanced evaluation offering.
On occasion, an NREC program demonstration leads to another project. Crusher was developed as a vehicle to conduct UPI experiments and tests. Personnel at the NREC took the requirements provided by the Defense Advanced Research Projects Agency (DARPA) for the UPI and started with a clean slate to invent a configuration that made sense for the program. They had to figure out how to take advantage of a vehicle with no cockpit as well as what it would look like. UPI personnel now are examining how to enable the vehicles to travel over terrain that soldiers traverse, to navigate intelligently and to conduct missions autonomously.
The NREC recently signed a new contract with the U.S. Army Tank-Automotive Research, Development and Engineering Center (TARDEC) to build an updated version of the Crusher as part of the Autonomous Platform Demonstrator program. TARDEC plans to add, on a work-directive basis, the requirement for NREC engineers to develop an unmanned ground vehicle end-to-end control architecture and to demonstrate the viability of the vehicle’s autonomous operations in a relevant environment.
When NREC staff members began their work with the UPI program, DiAntonio says they never intended to demonstrate how it would actually work using Future Combat Systems mission planning. With the new program contract, the Army will have the ability to evaluate all the infrastructure systems around the autonomous vehicle and obtain more information about how these types of robots could be used.
The NREC differentiates between its projects with unmanned vehicle design and its autonomous vehicle technologies work. Each category is a separate project area within the center. DiAntonio explains that the unmanned vehicles projects focus more on the design aspect, such as the best form for a nine-pound robot. The autonomous vehicle project area focuses more on the sensors and software necessary to make the vehicle intelligent. Platforms for those projects may be developed at the center, or personnel may take advantage of existing platforms. Other NREC project categories include operator-assist technologies, innovative mechanisms, sensing and image processing applications and machine learning applications.
Of all the programs under the NREC’s various categories, DiAntonio says that the UPI is the flagship. In February, the UPI team performed a field test at
|The Crusher UGCV was developed by the National Robotics Engineering Center (NREC) to support|
the UGCV Perception
for Off-Road Robotics (PerceptOR) Interface program’s experiments and tests.
In a more remote environment, such as a meadow, relying on sensors to make decisions based on geometry can result in failure. For example, a sensor might pick up on what it perceives to be a solid wall when the obstacle is actually tall grass. To overcome the problems, engineers must add color and other sophisticated algorithms so vehicles can determine when an object is, for example, grass that it can roll through. “UPI is without a doubt the most sophisticated robot for that type of terrain,” DiAntonio says.
Engineers with the program also are trying to take advantage of overhead imagery. The robots benefit from receiving satellite or other high-resolution imagery before they begin their missions because the engineers employ techniques to help base a path on the overhead information. Program officials have shown that the robots move faster and more efficiently using that type of imagery.
Another big effort at the NREC is DARPA’s Urban Challenge. In contrast to the UPI program, this project involves autonomous vehicles navigating through urban areas. “There, the challenge is dynamic obstacles,” DiAntonio states. The NREC did not take the lead on Carnegie Mellon’s Urban Challenge project but contributed resources and personnel.
The NREC has noticed several trends in the robotics industry, both government and commercial. In the military, DiAntonio says that one of the biggest trends is the comfort level and familiarity young officers and noncommissioned officers have with robotics. As they become tomorrow’s leaders, these officers will begin finding new ways to apply the technology.
The military’s next big step would be to move larger and faster robots into the field. One such project at the NREC is Gladiator. “That is what I would call a tactical robot, one that would actually go in front of infantry, Marine infantry in this case,” DiAntonio explains. The goal underpinning the various robot development programs—from current technology used to handle improvised explosive devices to the future devices that precede human forces—is to keep troops out of harm’s way.
In the commercial world, DiAntonio sees trends in the mining and farming industries toward using more robotics. Mining companies can use robots to find materials in locations too dangerous or impossible for humans to access, and farmers already are putting GPS infrastructure in place. The challenge on the commercial side is more a funding issue than a development issue because of the cost of many of the systems. In the near term, however, DiAntonio expects robotics to grow more in the private than public sector, as the technology becomes more affordable.
In addition to the current work at the center, officials at the NREC are looking at future trends. DiAntonio identifies one trend as learning techniques for robots. Engineers are developing techniques to study whether a robot can learn from its mistakes. “That is the technique I think that will dominate over the next 10 years,” he shares.
As engineers build the software, they need to solve the perception problem robots encounter. Humans have eyes as sensors and the ability to interpret what they see. “The difficulty in replicating that in hardware and computing is quite challenging,” DiAntonio states.
He also identifies other areas that could benefit from robotics such as convoying and homeland security. The U.S. Department of Homeland Security currently is examining the use of robotics in border control. Much of the technology can build off previous robotics investments.
UPI Program: www.rec.ri.cmu.edu/ projects/upi/index.htm
U.S. Army Tank-Automotive Research, Development and