Servicenavigation

Project Details

Back

Projekt Print View

Robust Algorithm Development and Demonstration for Agile mmWave MIMO Hybrid Beamforming

Subject Area Electronic Semiconductors, Components and Circuits, Integrated Systems, Sensor Technology, Theoretical Electrical Engineering
Term from 2019 to 2024
Project identifier Deutsche Forschungsgemeinschaft (DFG) - Project number 421544431
 
Final Report Year 2024

Final Report Abstract

Contribution of the Humboldt Universität: In the context of hybrid beamforming, knowledge of the geometry of the environment can be utilized very beneficially. Efficient beam configurations can be chosen on the basis of the geometric features of the environment without the need of lengthy training processes. The parameters for a mmWave beamforming system that supports fast beam-switching and Joint Communication and Sensing (JCAS) were defined. One main task was formalized as the room reconstruction problem that can be solved by performing full beam sweeping from transmitter and receiver side. The estimated channel impulse response information is used to find the reflection points, therefore, to determine the best possible beams for switching in case of decreasing quality of the wireless connection. Several algorithms that solve the room geometry reconstruction task were developed. The methods were developed using simulations based on ray-tracing channel models and verified experimentally using measurements in an anechoic chamber. The experiments and measurements show, that on the basis of angular resolved channel impulse responses (CIR), mayor objects in a room can be identified and located. Contribution of the Vodafone Chair, Technische Universität Dresden: A fully digital implementation of beamforming with AD/DA converters, mixers and other circuit components becomes infeasible for large antenna numbers at high frequencies due to the low area efficiency of a single antenna. For this reason a combination of analog and digital implementation and signal processing, known as hybrid beamforming, was investigated. As two PhD students were supported within the project at TU Dresden, two somewhat different topics but both related to digital and hybrid beamforming were pursued. The focus in the first part of the work was laid on developing efficient algorithms to select appropriate beams from a set of alternatives (usually denoted as a beamforming ‘codebook’ implemented with analog circuitry) to speed up the standard method of executing consecutively fixed length tests for all beams. To this end, three training schemes were investigated: 1) The previously proposed sequential test using a competition & elimination strategy to adapt the number of observations to the ambient SINR 2) The acquisition of prior knowledge (in the sense of a Bayesian approach to the decision problem) by Q-learning. 3) The idea of a frequency dependent beamformer (in contrast to the typically assumed narrowband assumption that claims that the beam pattern at all frequencies within the bandwidth is equivalent to the pattern at the center frequency fc) that is based on true time delays that are of the order of the symbol duration or the reciprocal bandwidth. This approach turns the unwanted effect of beam squint into an advantage, by creating a bijective mapping between the main lobe direction Φ of a uniform linear array and the frequency f ∈ [fc – fb /2, fc + fb /2]. As a second research direction the differences or potential equivalence between hybrid and digital beamforming was studied. Our findings indicate that for systems that are moderately loaded in the sense that the number of users is smaller or equal than the number of active beams (that in turn can be smaller or very much smaller than the number of antenna elements) both beamforming schemes are essentially equivalent. Intuitively, this can be understood by the fact that the weighted beam patterns (either by weighting all antenna signals or by weighting a set of selected beams) in both cases achieve the same amount of interference suppression as a consequence of the features of the beams such as main lobe width and number of roots that occur already in digital beamforming.

Publications

 
 

Additional Information

© 2025 DFG Disclaimer / Copyright / Privacy Policy

Textvergrößerung und Kontrastanpassung