The real-time knowledge of mmWave device orientation offers dual benefits for wireless networks and Internet of Things (IoT) applications: it enhances communication and localization performance through link diagnosis, and it enables context inference with asset and wearable tracking. This paper presents MetaPol, a novel system architecture that augments commodity mmWave access points (APs) with a carefully designed ultra-low-cost anisotropic metasurface to non-invasively extract the orientation of client devices, without hindering data reception or modifying hardware. MetaPol leverages the polarization of transmitted electromagnetic (EM) waves as an accurate indicator of the orientation of linearly polarized antenna arrays, prevalent in commercial mmWave mobile and IoT devices. Yet, polarization sensing is rarely supported by commodity APs due to the need for two orthogonally polarized antenna arrays to capture the incident electric field. Instead, MetaPol creates virtual polarization channels to characterize impinging electric fields, through the conversions of wave polarization on the fly as it interacts with the metasurface. To design MetaPol, we model and exploit the properties of a unique anisotropic metamaterial based on C-shaped split-ring resonators. We discover that, when grouped in certain configurations, these meta-elements can convert the incident polarization in a deterministic way. We show that three polarization channels are sufficient for unambiguous orientation sensing, and we provide a corresponding three-shot non-coherent protocol that extracts user orientation by comparing the power received from distinct surface areas. Through extensive over-the-air experiments with more than 1000 measurements in the mmWave band, we demonstrate that MetaPol achieves a mean error of 2.6° across practical settings with negligible compromise to the underlying data communication link.

Distinct from low-frequency analog design, design of radio-frequency, millimeter-wave and terahertz frequency chips follows a complex co-design process between circuits and electromagnetics (EM). These two domains are strongly coupled to each other, and this makes the design process extremely complex. Similar to circuit topologies, EM topologies can be very diverse ranging from lumped elements such as inductors, capacitors to distributed passive structures such as transmission lines, antennas, and all possible combinations, taking one or more elements of the set. The design process from ideation to schematic to layout can be extremely time-consuming relying heavily on expert knowledge, manual interventions, and iterative tuning of predefined circuit and EM topologies. These methods follow a bottom-up approach, starting with fixed geometric templates and using trial-and-error optimization. Other than long design times, this limits innovation and accessible design spaces. Here, we present a series of approaches that introduce a fundamentally different strategy: a universal inverse design method that can generate complex electromagnetic structures and subsequently fabrication-ready RFIC design and layout from specifications. We present a reinforcement learning design framework that canvasses the space for RFIC topology, circuits and circuit parameters and then synthesizes the interface electromagnetic structures through a deep learning enabled inverse design framework. Collaboratively, this creates a specifications-to-GDS design and synthesis flow for RFICs. The designs that emerge through this are often non-traditional and non-intuitive, but can break many of the trade-offs of known topologies. We present fabricated and measured results in silicon ICs demonstrating the feasibility of the presented approaches.

The design of radio-frequency reaching up to higher millimeter-wave (mmWave) and Terahertz (THz) frequencies is considered a black art. It requires years of human enterprise, deep human insights, co-design across multiple domains (such as circuits, layout and electromagnetics), and time/resource intensive iterative design simulations. The methodology is highly bottom - up using single functional elements, and subsequently building larger and more complex structures with it. Also unlike digital processors, there has not been a successful path towards automated design flow in RFIC yet. In this paper, we demonstrate a deep-learning based approach towards high frequency design that allows not only an algorithmic design approach, but also opens up a new design space that we have not explored before. We show inverse design of complex arbitrary shaped electromagnetic (EM) structures with designer S-parameters synthesized in minutes, which can be combined with active circuitry to realize circuits with unprecedented performance. We show this with a deep learning based mmWave PA with a record 30-94+ GHz Psat,3dBbandwidth, supporting concurrent multiband transmission for the first time at mmWave. We also demonstrate rapid synthesis of complex multi-port structures, allowing a path towards power combining with unique capabilities. We comment on the path towards full end-to-end RFIC design, synthesis and layout. AI can transform RFIC design when utilized effectively.

mmWave antenna array calibration is a necessary yet tedious and costly process in manufacturing to capture the non-idealities in phased arrays, in order to obtain codebooks for accurate and stable beam steering. Unfortunately, predefined codebooks provided by manufacturers to steer beams in a given set of directions do not support the arbitrary beam shapes required for various mmWave communication, sensing, and security applications. To create arbitrary beam patterns, one needs to first find the unknown calibration vector for the particular phased array in use. In this paper, we introduce EiCal, a novel zero-shot technique that leverages the beamforming codebook advertised by the manufacturer to extract the calibration vector at zero cost (i.e., with no additional measurements). The key idea is that the unknown desired calibration vector can be obtained via an appropriately designed eigen-decomposition of the given codebook. We experimentally demonstrate the efficacy of EiCal on a 60 GHz mmWave array for two scenarios: angle estimation using compressive pseudorandom beams, and simultaneous steering of beams and nulls. Our results also point to potential simplifications in the calibration process at the manufacturer.

Reconfigurable surfaces enable on-demand manipulation of electromagnetic wave properties in a controllable manner. These surfaces have been shown to enhance mmWave wireless networks in many ways, including blockage recovery. In this paper, we investigate the security vulnerabilities associated with the deployment of reconfigurable surfaces, i.e., an adversary may deploy new rogue surfaces or tamper with already-deployed surfaces to maliciously engineer the reflection pattern. In particular, we introduceMetasurface-enabled Sideband Steering (MeSS), a new metasurface-in-the-middle attack in which the spectral-spatial properties of the reflected wavefront are manipulated such that a concealed sideband channel is created in the spectral domain and steered toward the eavesdropper location, while maintaining the legitimate link toward the victim intact. We fabricate a custom reconfigurable surface prototype and evaluate MeSS through theoretical analysis as well as over-the-air experiments at the 60 GHz band. Our results indicate that MeSS significantly reduces empirical secrecy capacity (up to 81.7%) while leaving a small power penalty at the victim that can be masked under normal channel fluctuations.