Confidence Based Compressed Sensing Approach with Qos Differentiation for Reliable Grant-Free Access

Abstract

With the rapid development of 5G technology, numerous intelligent terminals can be successively connected to a network. This brings great challenges to existing network access because it increases the collision probability. It is very important for Grant-free random access (GF-RA) to ensure lowdelay and reliable transmission. Due to multiuser interference and varying environments, user detection can be extremely difficult, which leads to unreliable transmission. To address these limitations, research on the GF-RA method based on compressed sensing is carried out and includes the following: (1) The confidence-weighted orthogonal matching pursuit method is used for estimating the number of active users and overcomes the constraint of sparse condition. (2) The confidence-weighted simultaneous hard thresholding (SHT) method is used for active user detection (AUD), which can improve the performance of the SHT algorithm. (3) We conduct collision probability analysis to not only provide the performance upper bound using the multiple preambles method but also infer the collision probability relationship with different priorities. This is helpful to realize diverse quality of service requirements. The numerical results prove that the proposed method is valid for nonsparse signals and can realize a high success access rate.

Keywords: MIMO, Grant-free access, Compressed sensing, user detection, quality-of-services (QoS)