Coherent spatio-temporal sensor fusion on a hybrid multicore processor system

We report on the development, implementation, and demonstration of a novel, massively parallel computational scheme for detection of a target radiating a random signal in the presence of noise. This scheme involves coherent spatio-temporal fusion of data streams from multiple sensors, leading to the derivation of LLR detection statistics. Since streaming multicore processors with multi-SIMT architectures open unprecedented opportunities for fast signal processing, our algorithms are implemented on an NVIDIA Tesla C2050 many-core processor. Results achieved to date demonstrate up to two orders of magnitude speedup over a parallel implementation on a conventional quad-core processor, on a per-target-kinematic-hypothesis basis.