Abstract
The rapid proliferation of intelligent systems (e.g., fully autonomous vehicles) in today's society relies on sensors with low latency and computational effort. Yet current sensing systems ignore most available a priori knowledge, notably in the design of the hardware level, such that they fail to extract as much task‐relevant information per measurement as possible. Here, a “learned integrated sensing pipeline” (LISP), including in an end‐to‐end fashion both physical and processing layers, is shown to enable joint learning of optimal measurement strategies and a matching processing algorithm, making use of a priori knowledge on task, scene, and measurement constraints. Numerical results demonstrate accuracy improvements around 15% for object recognition tasks with limited numbers of measurements, using dynamic metasurface apertures capable of transceiving programmable microwave patterns. Moreover, it is concluded that the optimal learned microwave patterns are nonintuitive, underlining the importance of the LISP paradigm in current sensorization trends.
Highlights
Numerical results demonstrate accuracy improvements around 15% for object recognition tasks with limited numbers of measurements, using dynamic metasurface apertures capable of transceiving programmable microwave patterns
It is crucial in a variety of specific contexts—for instance, to limit radiation exposure, to adhere to strict timing constraints caused by radiation coherence or cluded that the optimal learned microwave patterns are nonintuitive, underunknown movements in a biological conlining the importance of the learned integrated sensing pipeline” (LISP) paradigm in current sensorization trends
Most attempts at task-specific tailored illuminations seek to synthesize illumination wavefronts matching the expected principal components of a scene.[20–23]. These principal component analysis (PCA) based approaches can be interpreted as a step toward optimal wave-based sensing: they incorporate a priori knowledge about the scene but ignore a priori knowledge about measurement constraints and the task at hand
Summary
The simplest approach in terms of the transceiver hardware is often to use random illuminations, for instance, by leveraging the natural mode diversity available in wave-chaotic or multiply scattering systems.[12–15] Random illuminations have a finite overlap that reduces the amount of information that can be extracted per additional measurement. A truly orthogonal illumination basis, such as the Hadamard basis,[16–19] overcomes this (minor) issue, often at the cost of more complicated hardware These “generic” illuminations fail to efficiently highlight salient features for task-specific sensing, which is necessary to reduce the number of required measurements. In other words, they do not discriminate between relevant and irrelevant information for the task at hand. Most attempts at task-specific tailored illuminations seek to synthesize illumination wavefronts matching the expected principal components of a scene.[20–23] These principal component analysis (PCA) based approaches can be interpreted as a step toward optimal wave-based sensing: they incorporate a priori knowledge about the scene but ignore a priori knowledge about measurement constraints and the task at hand (see Figure S1 in the Supporting Information). By training the ANN with a standard supervised learning procedure, we can simultaneously determine optimal illumination settings to encode relevant scene information, along with a matched postprocessing algorithm to extract this information from each measurement—automatically taking into account any constraints on transceiver tuning, coupling and the number of allowed measurements
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
