Student Topics

The multidimensional characterization of the radio channel and models derived from it are a core component in the development of radio systems for newly accessible frequency ranges. An important focus of current research is the (sub-)THz frequency range, which promises high data rates and low latency for future mobile communications standards (6G)¹. Due to significant changes compared to the familiar sub-6 GHz or mm-wave bands, metrology, i.e., the verification and characterization of measurement equipment, is also an important research topic that must deliver concepts for the calibration and certification of measurement technology in parallel with the actual applications. As part of the DFG-funded research group “METERACOM – Metrology for THz Communication,” the EMS group is developing metrological concepts for the multidimensional characterization of the radio channel.

The multidimensional characterization of the radio channel and models derived from it are a core component in the development of radio systems for newly accessible frequency ranges. An important focus of current research is the (sub-)THz frequency range, which promises high data rates and low latency for future mobile communications standards (6G)¹. Due to significant changes compared to the familiar sub-6 GHz or mm-wave bands, metrology, i.e., the verification and characterization of measurement equipment, is also an important research topic that must deliver concepts for the calibration and certification of measurement technology in parallel with the actual applications. As part of the DFG-funded research group “METERACOM – Metrology for THz Communication,” the EMS group is developing metrological concepts for the multidimensional characterization of the radio channel.

At EMS we conduct extensive radio measurement campaigns, which are the basis for future radio standards. This measurement data boils down to gigabytes per second of raw data collected over hours of measurement campaigns ¹. Afterwards, we can compute the estimated radio channel parameters from those measurements. A major bottleneck is the interpolation of the measured antenna characteristic ². Hence, we wish to improve compute throughput by using vectorized computation. This research project will give insights to how our existing algorithms could be vectorized on the tu ilmenau cpu hpc cluster.

Deep Learning methods have shown that they can be efficiently used to estimate radar parameters on synthetic datasets that were used both for training and performance evaluation^{1 2}. However, applying these methods to real measurement data remains challenging. This is mostly due to the scarcity of labeled measurement data and the associated difficulty of obtaining sufficiently labeled datasets³.

For first and second-order optimization methods, one usually is in need of two things. A step direction and a suitable stepsize to allow rapid convergence. Usually one employs a condition in the form of ¹ or uses a so-called Trust-region² to limit the stepsize during the iteration. If we look at it from another standpoint, we can consider for example a set of various different step directions and we want to find a suitable weighting of those in order to minimize the cost-function as rapidly as possible. Finding these weighting factors is the job of a step direction selection method.

When making use of the Compressed Sensing (CS) paradigm, one has to design the measurement process by means of a suitable compression step. Traditionally, the compression is modelled by means of a linear mapping acting on discretized version of the encountered signals, i.e. a usually complex-valued matrix ^{1 2 3}. Also, it has been an active field of research to apply CS as a data reduction scheme in radio channel sounding ⁴. Afterwards one is confronted with the problem to recover the parameters of interest, like time-of-flight, direction-of-arrival or Doppler-shifts from spectral-, spatial- and temporal measurements of a radio channel. In this context the iterative maximum-likelihood approach in ⁴ presents an optimization approach that is both computationally feasible and reproduces closely what is dictated by the Cramer-Rao-Lower-Bound (CRLB).

It has been observed in ¹ and Chapter 2.5 of ² that in channel estimation and modelling a substantial proportion of the transmit energy is not resolveable by a superposition of specular paths, which mostly adhere to a ray of propagation model. These so called diffuse multipath components (DMC) which do not follow the specular ray model pose two challenges.

It has been shown¹ that Linear Feedback Shift Registers (LFSR) can serve as a Radar sensor by random demodulation that enables significantly lower data-rates than dictated by the Nyquist-Sampling Theorem. At the same time they can be implemented very efficiently in silicon, while still allowing high configurability. This makes this approach a prime candidate for a wide range of Radar and inspection applications, but it requires a thorough study of the resulting theoretically achieveable performance.

Electromagnetic Raytracing¹ has become much more relevant for future digital twin modelling, as it now supports differentiation of the raytracing output. Not only does this allow smooth interpolation of the raytracing result, for instance in space and time, but also the integration of raytracing into learning architectures, which in turn solve inverse problems. This term is coined physics based deep learning².

The ESPARGOS development board¹ allows easy access to IQ data obtained from Wifi communications. As such, it efficiently sniffes an environment while providing realtime access to pilot symbols and respective channel state information. This allows to use of parameter estimation techniques based on maximum likelihood like².