Low-Complexity Angle-of-Arrival Estimation In Millimeter-Wave Large-Scale Hybrid Arrays

Millimeter-wave(mm Wave)frequencies and massive multiple-input and multiple-output(MIMO)have been accepted as the enabling technologies of the fifth-generation mobile communication(5G).On one hand,the high beamforming gain of the massive MIMO array can compensate the severe path loss of mm Wave frequencies;and on the other,the small wavelength of the mm Wave frequencies can accommodate hundreds to thousands of antennas in a limited space.The fully digital implementation of massive MIMO arrays,one radio frequency(RF)chain per antenna,can be impractical for the majority of commercial applications,due to the large power consumption and high cost of the mm Wave RF chain and the small size of the mm Wave array.To fully unleash the potential of the mm Wave massive MIMO in 5G,the large-scale hybrid analog/digital array(LHA)has been recognized as a key technology.Two LHAs,known as the large-scale array of phase shifter-based subarrays(LAPSS)and the large-scale array of Butler matrices(LABM),have attracted extensive attention,due to their low complexity,high reconfigurability and the robustness against malfunction.The accurate angle-of-arrival(AoA)estimation is the basis of translating the LHAs into effective applications.However,the few existing methods specifically designed for the LAPSSs and LABMs fail to estimate the AoA fast and accurately.Some of them can have the prohibitive computational complexity;some need to iteratively estimate the AoA,leading to the lengthy training delay and the error propagation;and the others have the estimation ambiguity problem.Against the above background,this dissertation studies the AoA estimation in LAPSSs and LABMs.The key contributions are summarized as follows.A non-iterative and low-complexity AoA estimation method is proposed for the LAPSS.The method first estimates the propagation phase difference between adjacent subarrays,denoted by Nu,and suppress the phase differences to extract the AoA.The main contributions include:(1)a new analog beamforming scheme is designed for the unambiguous AoA estimation and eliminates the constraints on the parameters of hybrid arrays in the existing works;(2)it is unprecedentedly discovered that the signs of the cross-correlations between consecutive subarray outputs conform to a deterministic rule and the rule has a fixed pattern across symbols;(3)an unambiguous method for the estimation of Nu is developed by exploiting the new discoveries;(4)it is proved that the AoA conforms to a one-to-one mapping to the spatial responses of the selected subarrays,and the mapping can be applied to all the subarrays available;(5)the mean squared error lower bounds(MSELBs)of the proposed Nu and AoA estimates are derived.Corroborated by extensive simulations,the proposed method can reduce the mean squared error(MSE)of the AoA by as much as 86%,as compared to the state-of-the-art methods.A low-complexity and accurate wideband AoA estimation method is proposed for the wideband LAPSS,which adopts the same analog beamforming scheme as developed for the narrowband scenario.The main contributions include:(1)it is proposed to estimate Nu from the second cross-correlations between the cross-correlations of any consecutive subarray outputs at different sub-carriers;(2)the sub-carrier interval is derived to attain the optimal trade-off between the accumulation gain and the error amplification when estimating Nu;(3)it is proved that the sign alternating rule and pattern of the subarray cross-correlations,and the one-to-one mapping for AoA estimation,discovered in the narrowband scenario,can be applicable to any sub-carrier in the wideband scenario;(4)it is proved that the signal of interest can be coherently accumulated over sub-carriers with the maximum accumulation gain achieved;(5)the MSELB of the proposed wideband AoA estimate is derived.Extensive simulation results validate the proposed method and illustrate that the AoA estimate can be improved by up to 90 %,as compared to the state of the art.An expeditious AoA estimation method for the LABM is proposed,which first estimatesNu between the consecutive Butler matrices and then extracts the AoA from the judiciously selected Butler beams.The main contributions include:(1)a new wide beam synthesis method is proposed for the Butler matrix,which efficiently synthesizes the Butler beams in the analog domain through beam selection;and the synthesized wide beam has the flat mainlobe and the large mainlobe-to-sidelobe ratio,and can be reconfigured flexibly in terms of the pointing direction and the beam-width;(2)by exploiting the new wide beam synthesis method,a fast search of the angular region of the impinging path is achieved,and as a result,the same Butler beams are selected across the Butler matrices,and the high beamforming gain is exploited for Nu estimation,leading to the high-accuracy Nu estimate;(3)it is proved that by judiciously selecting a set of Butler beams,the AoA can be uniquely extracted in a closed-form expression;(4)the noise resistance of the AoA estimate is analyzed,disclosing that only the selected set of Butler beams dominate the AoA estimation performance;(5)the MSELB of the proposed AoA estimate is derived.Extensive simulations validate the superiority of the proposed method over the state of the art;and specifically,by translating the AoA estimates into the system achievable rates,up to 70 % rate improvement can be achieved by the proposed method.Another accurate AoA estimation method is proposed for LABMs,which exploits the am-plitude information of Butler beams.The main contributions include:(1)by analyzing and exploiting the simultaneous multi-beam steering ability of the Butler matrix,the differential Butler beam(DBB)is designed and is proved to have the higher received signal-to-noise ratio and,in turn,the detection probability,than those of the original Butler beams;(2)it is proved that the AoA can be estimated from the first two strongest DBBs;and a closed-form solution is derived to extract the AoA from the strongest DBBs;(3)the new wide beam synthesis method is exploited to achieve a fast search of the first two strongest DBBs,which results in the logarithmic searching time in contrast to the linear time complexity of the state of the art;(4)the MSELB of the AoA estimate is derived.Corroborated by extensive simulations,the proposed method can achieve the performance improvement by up to 50 %,90% and 64 %,in terms of the detection probability of the DBBs,the AoA estimation accuracy and the time-average beamforming gain,respectively,as compared to the state of the art.