Compact Aerial Video Exploitation (CAVE)

Small Unmanned Aerial Vehicles have become a critical part of Intelligence, Surveillance and Reconnaissance (ISR) missions, supplying valuable aerial imagery to ground forces. Unfortunately, operational ISR is compromised by the stringent size, weight, and power constraints of small UAVs to support real-time processing of imagery and transmitting valuable ISR results over limited-bandwidth communication links. Therefore, the potentially game-changing combination of high-resolution ISR and tactical-edge availability has not yet materialized. Our proposed solution is an integrated hardware/software system designed to process high-resolution video data at full video-rate onboard a small UAV. The software performs target recognition and feature-aided tracking, such that only narrow-bandwidth results need to be transmitted back to the user. Our solution employs a revolutionary computing hardware architecture offering several orders of magnitude greater efficiency over conventional processors in terms of throughput vs. power consumption. The key innovation is a slightly reduced precision arithmetic logic unit (ALU) built from just a few thousand transistors, instead of the hundreds of thousands used in modern floating point units. Our simulations show that simple parallel architectures based on these extremely small ALUs run applications 10,000 times more efficiently (faster, or lower power) than modern CPUs and 100 times more efficiently than GPUs.