Ghost modes—false positives produced by RF demodulators—carry an operational cost.They waste computational resources, trigger spurious alarms and can mask true signals. Inreal-time systems, where performance criteria must be met every cycle[2], ghost hits competewith latency…
Exhaustive parameter sweeps are the de facto method for benchmarking RF pipelines, butthey scale poorly with dimensionality[2]. For example, a 10-parameter grid with 10 points eachrequires 1010 evaluations. Active learning promises to achieve comparable confidence…
Exhaustive parameter sweeps are the de facto method for benchmarking RF pipelines, butthey scale poorly with dimensionality[2]. For example, a 10-parameter grid with 10 points eachrequires 1010 evaluations. Active learning promises to achieve comparable confidence…
Accurately characterising the performance of black-box RF demodulation pipelines normallyrequires dense sweeps over many parameters, incurring high computational cost[1]. Gaussianprocess (GP) surrogate models provide smooth interpolants and predictive uncertainty, enablingefficient exploration[1]. This paper investigates how…
Real-time RF inference systems must meet strict latency budgets to be considered operational [2]. Latency is not uniform across the multidimensional parameter space: frequencyoffsets, signal-to-noise ratios and other demodulator settings can trigger sudden spikes in…
In RF demodulation pipelines, system failure is often characterised by the emergence ofghost modes and a collapse of true detections. These “failure rims” occupy narrow regionsof a high-dimensional parameter space spanned by signal-to-noise ratio (SNR),…
Robust radio-frequency (RF) pipelines must balance conflicting objectives: accurate demodulation of signals and low latency. Improving accuracy (true hit rate) often requires operatingin regimes with high signal-to-noise ratio (SNR) and small frequency offset, which may…
Exploring the configuration space of radio frequency (RF) demodulation pipelines to find robust operating regimesis challenging because simulation runs are computationally expensive and the response surface can contain sharpfailure boundaries. A naïve uniform grid of…
RF-augmented reality (RF-AR) wearables enabledetection of threats, casualties and anomalies, but they rely ona set of alert thresholds tailored to mission context and userphysiology. Presently, these thresholds are either hard codedor tuned manually, limiting adaptability…
Augmented reality (AR) headsets promise to improve situational awareness for medics, firefighters and securitypersonnel by overlaying live information on the physical world.However, existing AR user interfaces struggle when many targets must be tracked simultaneously: cluttered…