Riding the Wave with LoFLYTE

The Low-Observable Flight Test Experiment (LoFLYTE) program was a joint effort among researchers at NASA Langley and the Air Force Research Laboratory with support from NASA Dryden and the 445th Flight Test Squadron at Edwards Air Force Base. Accurate Automation, Corp., of Chattanooga, TN, received a contract under NASA’s Small Business Innovation Research program to explore concepts for a stealthy hypersonic wave rider aircraft. The Navy and the National Science Foundation provided additional funding. A wave rider derives lift and experiences reduced drag because of the effects of riding its bow shock wave. Applications for wave rider technology include transatmospheric vehicles, high-speed passenger transports, missiles, and military aircraft.

The LoFLYTE vehicle was designed to serve as a testbed for a vari­ety of emerging aerospace technologies. These included rapid prototyp­ing, instrumentation, fault diagnosis and isolation techniques, real-time data acquisition and control, miniature telemetry systems, optimum antenna placement, electromagnetic interference minimization, advanced exhaust nozzle concepts, trajectory control techniques, advanced land­ing concepts, free-floating wingtip ailerons (called tiperons), and adap­tive compensation for pilot-induced oscillations.[1008] Most important of all, LoFLYTE was eventually to be equipped with neural network flight controls. Such a system employs a network of control nodes that inter­act in a similar fashion to neurons in the human brain. The network "learns,” altering the aircraft’s flight controls to optimize performance and take pilot responses into consideration. This would be particularly useful in situations in which a pilot needed to make decisions quickly and land a damaged aircraft safely, even if its controls are partially destroyed. Researchers also expected that neural network controls would be useful for flying unstable configurations, such as those necessary for efficient hypersonic-flight vehicles. The computing power of Accurate Automation’s neural network was provided by 16,000 parallel neurons making 1 billion decisions per second, giving it the capability to adjust to changing flight conditions faster than could a human pilot.[1009] The LoFLYTE model was just 100 inches long, with a span of 62 inches and a height of 24 inches. It weighed 80 pounds and was configured as a narrow delta planform with two vertical stabilizer fins. The shell of the model, made from fiberglass, foam, and balsa wood, was constructed at Mississippi State University’s Raspet Flight Research Laboratory and then shipped to SWB Turbines in Appleton, WI, for installation of radio control equipment and a 42-pound-thrust microturbine engine.[1010] The first flight took place at Mojave Airport, CA, on December 16, 1996. The vehicle was not yet equipped with a neural network and relied instead on conventional computerized stabilization and control systems. All went well as the LoFLYTE climbed to an altitude of about 150 feet and the pilot began a 180-degree turn. At that point—about 34 seconds into the flight—the ground pilot was forced to land the craft wheels-up in the sand beside the runway because of control difficulties. The model suffered only minor damage, and researchers generally considered the flight a suc­cess because it was the first time a wave-rider-concept vehicle had taken off under its own power.[1011] Testing resumed in June 1997 with several flights from the Edwards North Base runway. This gave researchers the opportunity to verify the subsonic airworthiness of the wave rider shape and analyze basic handling characteristics. The results showed that a full – scale vehicle would be capable of taking off and landing at normal speeds (i. e., those comparable to such high-speed aircraft as the SR-71). Flight tests of the neural network control system began in December 1997 and continued into 1998. These included experiments to verify the system’s ability to handle changes in airframe configuration (such as removal of vertical stabilizers) and simulated damage to control surfaces.[1012]