National Defense Magazine
Featured Article, Sep 2002

Ground Robots Experience Bumpy Ride

Top technology challenges are mobility, autonomy and communications

by Roxana Tiron

In the FCS, the unmanned vehicles are supposed to operate in forward areas, into enemy terrain. However, robotic systems currently lack the “robustness and flexibility” required to take on this role, according to the Defense Department’s 2002 Joint Robotics Program Master Plan.

A Pentagon official who is familiar with the robotics program said that semi-autonomous and autonomous mobility are still not feasible, but that “serious progress” could be achieved in about five years. “It is a lot closer than we thought it would be,” he said. “Responsive robotic dogs exist already.”

Current robots are operated by remote control (tele-operated). They do not think independently.

“When I am on the screen and give the system a command and the machine knows how to get from point A to point B—this is semi-autonomous,” said the official. “It’s like riding a horse. You don’t tell the horse to pick up its feet—the horse goes where you direct it.”

With tele-operation, there is a communication link between the operator and the platform “and that requires a lot of bandwidth. Ever so often you lose the link you hit a dead zone. And so, you can’t get the images back,” he said.

A semi-autonomous system would have sensors for collision or obstacle avoidance, even when the link is broken, he said. “You just send a command to it and the system is smart enough to go to the coordinate. so you don’t have to have the pipeline open. You help your communication problems tremendously.”

Tele-operated machines communicate either through RF (radio frequency) or through fiber optics—“which can be a tail to the machine and somebody can easily cut it,” he said.

There are specific missions where the communications links are vital, such as explosive ordnance disposal or hazmat (hazardous materials) operations. “The machine is an extension of [the operators] arms, ears and eyes,” he said.

Autonomous ground robots will have to have a real-time processing capability. It will have to think like a human, so to speak, the official said.

Such autonomy is what officials hope will bring advanced capabilities to the Future Combat System, particularly in roles such as intelligence gathering and re-supply missions.

The Army has put FCS on a fast track and is trying to have an initial operational capability before the end of the decade. To meet this ambitious timetable, the Army has adopted a two-tier approach. The initial fielding of FCS is slated to use unmanned systems with limited autonomous mobility capabilities. The technology will rely on human intelligence aboard the manned leader vehicle to “reduce the perceptual and intelligent control requirements for the unmanned follower vehicle,” according to the master plan.

For the long term, the goal is to boost the onboard intelligence of unmanned systems, thus reducing the reliance on human intervention to complete a mission.

The leader-follower concept for robot operations has been the source of some debate in the Army. Some experts question the wisdom of having multiple vehicles depend on a single leader. As one official put it, “all the enemy has to do is shoot down the leader vehicle, the one with all the antennas.”

Joint Robotics Program

The Joint Robotics Program, established in 1990, maintains those two tracks in its acquisition strategy: to develop and field first-generation unmanned ground vehicles with current technologies (tele-operation) and to pursue technologies for semi-autonomous mobility that gradually can be inserted into first-generation systems.

In 2002, the Pentagon’s ground robotics program received $27.6 million and a congressional plusup of $6.4 million.

According to the master plan, tele-operation is a proven technology and will be fielded with the first-generation UGVs. However, current tele-operated machines need to be improved, so they can perform multiple missions, said the report.

The first-generation UGVs include the Remote Ordnance Neutralization System (RONS), Standardized Robotic System (SRS), Robotic Combat Support System (RCSS) and All-purpose Robotic Transport System (ARTS).

Operation Enduring Freedom prompted the U.S. Air Force to order 30 RONS. The Air Force is also using ARTS overseas for force protection and homeland defense. Eighteen units have been fielded and the service has another order for 28 units.

Two years ago, the Army established the Semi-Autonomous Robotics for FCS Science and Technology Objective. The STO is scheduled to conclude in fiscal year 2005 and will focus on the development of mobility technology.

The STO will address the perception and intelligence required to allow vehicles to roam throughout the battle space, without “substantial” operator intervention, said the master plan. The technology has to be adaptable to various environments and missions.

The STO “will focus upon developing vehicle behaviors, analogous to the skills possessed by soldiers that will enable unmanned FCS elements to possess a reasonable level of tactical knowledge and adaptability,” said the report. Ideally, systems will ultimately be able to operate almost autonomously, employing “commander’s intent” to develop operational plans.

The STO incorporated technology developed under an existing Army-Defense Department program called Demo III, designed to look at both active and passive sensor models across the electromagnetic spectrum. These include radar, LADAR (laser radar), visible electro-optic and infrared imaging for the detection and classification of objects that may affect the vehicle’s mobility.

Since all unmanned systems will, for the foreseeable future, have soldiers in the loop, the program is also seeking to develop modular interface software for the control of multiple unmanned systems—which could be incorporated into future tactical vehicle command-and-control systems.

The current Demo III prototype is a four-wheel drive, four-wheel steer, hydro-static drive vehicle, weighing 3,200 pounds. Its engine is off-the-shelf and the vehicle has stereo sensors, FLIRs (forward looking infrared) and cold FLIRs, said Chuck Shoemaker, from the Army Research Lab’s Demo III program.

Other participants in the Demo III program include the Energy Department’s national laboratories, the National Institute of Standards and Technology, the ARL Robotics Collaborative Technology Alliance—a consortium led by General Dynamics Robotic Systems, with Carnegie Mellon University Robotics Institute, SRI International, Sarnoff, Jet Propulsion Laboratories and the University of Maryland, Applied Systems Intelligence, Micro-Analysis and Design, BAE Systems and Florida A&M University.

Demo III is under a 36-month contract worth $11.5 million, according to the ARL.

Research for Demo III has concentrated on three areas: perception, intelligent control and behaviors and man-machine interfaces.

Soldiers are brought into the development process through virtual and live experiments. Demo III is meant to develop and integrate technology that will enable a single soldier to operate of up to four unmanned vehicles while they maneuver off-road at speeds of up to 20 mph and on-road at speeds of up to 40 mph. The actual goal is to have vehicles go at half the speed of a manned Humvee truck in the same terrain.

According to the master plan, the program made significant strides in November, at field exercises conducted with soldiers from the 28th Infantry Division (Pennsylvania National Guard). Four experimental UGVs participated—three configured for reconnaissance, surveillance and target acquisition and one configured for obscurant dispersal functions.

These vehicles maneuvered autonomously throughout the terrain requiring only minimal operator intervention when they became “confused and were unable to rapidly proceed with the mission due to the relative myopia of the perception sub-systems,” said the master plan.

Because of the increasing interest in UGVs for the FCS and the Army’s Objective Force, the Defense Department agreed to extend the Demo III program for approximately 18 months beyond its initially scheduled conclusion in fiscal year 2002.

Over the next year, Demo III is expected to participate in the FCS Lead Systems Integrator Unmanned Combat Demonstration and conduct additional field exercises, said the ARL.

The program will also focus on basic behaviors such as seeking cover and concealment in rolling, forested terrain.

Subsequently, the program will work on the development of complex military behaviors and “while it is expected that they will be fairly brittle, it will provide the experience base required for the creation of adaptive behaviors later in the program,” said the report.

A key technology for creating such behaviors will be long-range perception algorithms and sensor suites that can perceive three-dimensional environments from as far as 500 meters away.

In the foreseeable future, the Army would like to test the tactical behaviors that will allow an unmanned vehicle to maneuver over terrain that no manned vehicle has crossed before.

By October 2004, the Joint Robotics Program STO plans to demonstrate higher speed (35 mph) off-road mobility for FCS-scale vehicles and cooperative behaviors such as two vehicles providing support cover during a longer route.

Reaching these goals will require higher resolution local terrain information at longer ranges, improved sensor data fusion and use of contextual data to provide cues for negative obstacles.

By 2005, the program will seek to implement and demonstrate adaptive tactical behaviors. Adaptive behaviors only can be achieved through the development of appropriate learning algorithms, said the report.

Ground Combat Vehicle

Meanwhile, DARPA and the Army are jointly funding the FCS unmanned ground combat vehicle (UGCV) program, which is being managed by DARPA. This program looks at the performance gains of combat vehicles without the constraints of accommodating an onboard crew.

The primary metrics for the UGCV are endurance (14 days and 450 km off-road), obstacle negotiation and payload fraction. Airdrop-ability, robustness to crash, reliability, signature and cost come as secondary metrics.

The UCGV program completed 11 preliminary designs over nine months in 2001. In October four of the designs entered critical subsystem testing and detailed prototype design. Two of those designs are capable of 1,500 kg of payload. These payloads can range from marsupial robots to sensors to weapons systems to smaller air vehicles.

The program is also looking at UCGVs with a payload of 150 kg, which would serve as reconnaissance and surveillance vehicles, carrying self-protection and even armor in some cases, said Scott Fish, DARPA’s project manager for the UCGV.

In July, DARPA awarded $5.5 million contracts to Carnegie Mellon University and Lockheed Martin Missiles and Fire Control to build UGCV prototypes. Both designs use hybrid-electric drive trains. These prototypes will undergo testing in 2003 and 2004.

The Army is looking at UCGVs, said Fish, because the service needs increased deployability that would come from reduced platform mass and volume and from simplified air-drop. Also, high endurance and lower manning are desired features.

“Robots will eventually crash,” said Fish. Contractors need to think about that—local control and recovery need to be included in the planning, he said at an industry conference.

However, Fish cautioned, “Robots will be the lowest priority for re-supply. Manned aircraft will not land in hostile territory to deliver robots.”

Another jointly sponsored Army-DARPA program is the FCS perception for off-road robotics (PerceptOR). The project is studying several perception approaches for off-road robotic navigation.

“We are trying to gather a bunch of data and find out what robots can do,” said Fish. “Honestly, we don’t know what robots can do.”

In this program, he said, higher resolution data will soon be available and it is going to contribute to safer and faster robot operations.

The PerceptOR program outfitted eight Honda all-terrain vehicles with four different perception approaches over eight months, beginning in March 2001. Three teams were selected in December for Phase II, which includes four unrehearsed experiments in 2002 to highlight the strength and weaknesses of each approach. The companies leading the three teams in this phase are SAIC, Carnegie Mellon and General Dynamics.

However, said Fish, the program is facing some technical challenges, such as the registration of data from multiple sensors into a common world map; terrain and object classification under various lighting conditions; maintaining the location understanding when the GPS (Global Positioning System) is degraded; operating near a potential obstacle, backing up and turning in confined spaces and adapting to weather effects.

other articles...