LAUSR

In this paper, a novel compressed-sensing-based channel-estimation (CE) method is proposed to estimate the sparse and doubly selective air-to-ground wireless channel for drones. These drones are supposed to be utilized for data transmission in cellular networks as well as traffic monitoring. The proposed method utilizes time domain training sequence for CE. The Monte Carlo simulation results indicate that the proposed CE method performs better in high Doppler shift conditions compared with other recently proposed methods. We also present a modification of the proposed method for sparse channels where the location of multipath receptions does not change as fast as their amplitude and phase over time. Exploiting this fact, we separate estimation of the sparse channel’s nonzero elements locations from estimating their amplitudes and phases and therefore improve bandwidth efficiency and performance. The proposed methods are used in simulation of a traffic surveillance scenario where frames of a real video which were taken by a drone from a highway are used as the input to the communication system and the received and demodulated signal is analyzed for car detection using a faster region-based-convolutional-network algorithm. The computational-complexity and the bandwidth-efficiency of the CE methods are compared within this scenario. Simulation results by using normalized-mean-square-error and bit-error-rate calculations show that using the proposed CE methods increases the CE and data demodulation accuracy whilst compared with the least-square method, increases the bandwidth-efficiency up to 5% and reduces the computational-complexity by a factor of two or more.