Probabilistic RF-Assisted Camera Wake-Up through Self-Supervised Gaussian Process Regression

Abstract

Research on wireless sensors represents a continuously evolving technological domain thanks to their high potentialities: flexibility and scalability, fast and economical deployment, pervasiveness in industrial, civil and domestic contexts. However, the maintenance costs and the sensors reliability are strongly affected by the battery lifetime, which may limit their use and exploitation. In this paper we consider a scenario in which a wireless smart camera, equipped with a low-energy radio receiver, is used to visually detect a moving radio-emitting target. To preserve the camera lifetime, we design a probabilistic energy-aware controller that regulates the camera state. The radio signal strength at the receiver side is used to predict the target detectability, via selfsupervised Gaussian Process Regression combined with Recursive Bayesian Estimation. Both numerical and experimental results validate the proposed approach in terms of target detection accuracy and energy consumption.

Transmitter Discovery through Radio-Visual Probabilistic Active Sensing

“…The radio sampling time is set to T RF = 0.1 s, and the value of ν is set to 10, as a reasonable trade-off between typical real-life values and the ideal condition stated in (21). The underlying POD function is (23), designed according to real-life experiments [13]. During the POD learning phase, the target is supposed to move according to the stochastic linear model reported in Tab.…”

“…RaViPAS aims to combine the complementary benefits of RF signals with visual cues. The literature addressing radio-visual sensor fusion is still sparse and the RaViPAS framework is an open research field with methodological challenges and application opportunities: for example, energy-aware strategies are required to alternate the accurate but energy-harvesting camera measurements, with the rough but lighter radio observations [11], [13]. Furthermore, traditional RF-based solutions involve tiring human-labeled calibration processes [12].…”

“…Furthermore, traditional RF-based solutions involve tiring human-labeled calibration processes [12]. In this regard, the integration of accurate camera measurements in self-supervised calibration methods have been proven to be beneficial [11], [13]. Contribution -In this work, we propose a RaViPAS scheme to solve the transmitter discovery problem hereafter described (Fig.…”