by Mike Deschenes and Craig Sanderson
Unmanned aerial vehicles (UAVs) are a massive growth area in the defense sector. The military's effective use of UAVs in recent conflicts has highlighted their successes, which are vigorously driving UAV technology developments.
Designers of UAVs are utilizing compact field-programmable gate-array (FPGA)–based systems to supplant conventional processors because of the demanding real-time processing requirements of many new UAVs under consideration, as well as the extraordinary size, weight, and power (SWAP) constraints inherent to UAV design.
These ultra-high-performance FPGA solutions deliver tens of billions of floating-point operations per second (giga FLOPS) at a fraction of the SWAP budget required by equivalent systems using conventional processors.
UAVs have transcended their traditional intelligence surveillance and reconnaissance (ISR) role and are now filling a much broader range of missions. Military agencies have extended their view of the UAV's role to include new mission types such as armed reconnaissance, strike and suppression of enemy air defenses (SEAD), air-to-air combat, and homeland security.
In addition, UAVs hold potential for use beyond the normal scope of military operations. Placed in the hands of the appropriate domestic agencies, UAVs could play a key role in law enforcement, border control, and Coast Guard duties.
These new missions, applications, and expectations present UAV developers with a significant range of challenges. In addition to incorporating an entirely new set of functions, next-generation UAVs need to make a leap forward in capability to perform in their new roles. Developers must address issues such as operation within civil airspace, vehicle maneuverability, and other logistical concerns.
As UAVs gain capability and intelligence, the sensor, communications, and computing technology in these vehicles must perform at a high level to keep pace with requirements. In addition, the significant SWAP restrictions that UAVs impose present a challenge to developers. This applies in particular to UAVs' extremely demanding computing technology, which is forcing developers to consider new approaches.
At the same time, UAVs are going mainstream, and the race is on for developers and technology to keep up. The U.S. Department of Defense (DOD) is expected to increase development funding for UAVs by 775 percent between 2000 and 2010.
Detect and avoid
A prime example of a next-generation UAV application is detect-and-avoid (DAA) systems, which are designed to perform many of the safety functions that would normally be performed by human pilots onboard an aircraft. These systems are particularly important when UAVs fly in civilian airspace where avoiding collisions with other vehicles is a prime concern.
The civil-aviation authorities of most nations require UAVs to operate with the same level of safety as manned aircraft. The U.S. Federal Aviation Authority (FAA) recognizes UAVs as remotely operated aircraft (ROA). The importance of this issue is recognized in the "DOD UAV Roadmap 2002–2027," which states, "see-and-avoid is a key issue in allowing ROAs into civilian airspace." (This report refers to DAA as "see and avoid.")
A DAA system must first detect air traffic and then determine whether the aircraft is on a collision course with the host UAV. Some UAVs already have the capability to detect aircraft with the use of transponder-based collision-avoidance systems. However, these systems only work if the other aircraft in the airspace are equipped with a compatible system, making a transponder-based system unsuitable for universal application in multiple civil airspaces — a requirement of many current and future UAV programs. The technology of DAA systems must evolve to meet these requirements demanded by the FAA and other civil-aviation authorities.
Current UAVs do not have onboard capability to detect other aircraft without transponder-based systems. Potential technical approaches include active systems, such as radars, but these are expensive in terms of SWAP, making them incompatible for most UAVs. Alternative technical approaches being developed operate autonomously and focus on passively detecting approaching air traffic to enable the UAV to avoid collisions. The complexity of such systems requires the use of high-performance onboard data processing to make the system effective.
Advanced applications such as DAA place a significant demand on a UAV's onboard processing capabilities. In acknowledgment of this fact, the "DOD UAV Roadmap 2002–2027" notes that "increased onboard processing will be the key enabler of more-responsive flight-control systems, onboard sensor data processing, and autonomous operations (AO) for future UAVs." Processing the high-bandwidth data associated with UAV applications necessitates performance levels of tens of giga FLOPS, particularly because these systems must often process the data in real time.
This type of processing is demanding and would normally require multiple-microprocessor supercomputing solutions. Because of the SWAP requirements in UAVs, such a solution is simply unfeasible. The payload of the General Atomics Predator UAV is 450 pounds and the payload volume is also restricted because of the small size of the vehicle — a situation that most other UAVs share.
Numerous flight and mission systems compete for this capacity, along with a finite onboard electrical power resource. This makes squeezing in as much processing capability and minimizing SWAP requirements a must for applications that make it onto UAVs. Using standard microprocessors and platforms to achieve these processing requirements would consume a significant amount of space, weight, and power beyond the availability of these commodities in most UAVs.
Consequently, developers need to look for different approaches to meet their processing needs. FPGA computing technology can deliver very high per-device processing capabilities. By exploiting the parallel processing capabilities made possible by FPGA application design, developers can extract multiple-giga-FLOP performance levels from one device. Compared to microprocessors, FPGAs provide higher processing capability relative to size, weight, and power consumed, making them attractive for UAV applications.
The other factor driving the uptake of FPGA computing in UAV applications is the availability of standard-product system solutions. In the past, FPGAs were considered excessively complex to implement for relatively low-volume applications such as UAVs. However, there has been significant growth in the availability of COTS (commercial off-the-shelf) FPGA computing solutions and design services, along with improvements in FPGA design-tool capability.
DAA case study
Defense Research Associates Inc. (DRA) in Dayton, Ohio, is developing a DAA system to meet all these requirements. Working with the U.S. Air Force Research Laboratory Sensors Directorate under the sponsorship of the Global Hawk and Predator program offices, the company's DAA system uses a low-cost, low-power approach based on high-resolution imaging sensors with proprietary detection and tracking algorithms.
The system uses complex and computationally demanding algorithms, through which it processes all of the video data from the high-resolution sensors. Unable to achieve the necessary performance levels using conventional processors, DRA turned to FPGAs. DRA partnered with Nallatech Inc. in Orlando, Fla., to implement the algorithms for operation in real time.
The FPGA hardware sits alongside conventional microprocessors, which DRA uses for less-demanding aspects of the DAA system such as system control and user interfacing.
In a recent series of flight trials, a single-sensor DAA demonstration system proved itself more than capable of meeting the DAA performance requirements of Global Hawk and Predator. The demonstration system, built on PC/104 form-factor boards that fit in a space of less that 5 inches square and 7 inches high, exhibited impressive performance figures, processing data at 36 giga FLOPS within the FPGAs.
This impressive performance density equates to delivering 60 times the real-time processing performance of a server-class microprocessor system, while satisfying the demanding SWAP constraints of UAV platforms. DRA is extending the capabilities of this prototype in preparation for flight on a UAV in the (U.S.) National Airspace System in 2005.
Increasing demands and budgets are rapidly fueling the growth of UAV development. The advanced, high-performance applications required to enable the increasing scope of UAV missions are driving the onboard processing performance and SWAP demands of UAV applications, pushing past the capabilities of microprocessor technologies.
As it becomes increasingly difficult to meet these needs using microprocessors, application developers of current and next-generation UAVs are already looking for alternative implementation technologies for onboard processing.
The performance density and low power consumption inherent to FPGA computing systems as compared to conventional processing systems uniquely qualifies this approach to meet the demands of UAV developers. FPGA computing solutions are quite simply the technology of choice for onboard processing in next-generation UAV applications.
Craig Sanderson is a systems application engineer for Nallatech Inc. in Orlando, Fla., who provides specification, support, and project management of application development. He also contributes to the specification, definition, and marketing/sales of new Nallatech products.
Mike Deschenes is senior engineer for Defense Research Associates in Dayton, Ohio, and is lead engineer working on detect-and-avoid system development, with responsibility for program management, requirements definition, system specifications, and system integration.