In this paper we study feedback mechanisms for unsourced random access (URA) communications. We propose an algorithm to construct feedback packets broadcasted to the users by the base station (BS) and a feedback packet format that allows the users to estimate their channels and infer positive or negative feedback based on the presented thresholding algorithms. We show that the proposed feedback technique leads to a substantial reduction in the packet error rates and signal-to-noise ratios (SNRs) required to support various numbers of active users. We also demonstrate that the proposed feedback imposes a much smaller complexity burden on the users compared to the feedback that acknowledges only successful or only all undecoded users. Finally, we present a theoretical analysis framework that closely matches the experimental results.

We focus on an unsourced random access (URA) system for communication over fading channels where the payload of each packet is encoded for error-correction, repeated, permuted, and scrambled. Each packet is also equipped with a preamble that is used for channel estimation and detection of permutation and scrambling sequences utilized for payload encoding. We propose an algorithm to resolve multiple-access preamble transmission, based on the approximate message-passing (AMP), that is capable to support high numbers of active users and achieve low probabilities of miss-detection. We also develop a parallel interference cancellation technique for payload reception that iteratively refines the channel estimates and attempts to minimize the mean squared error (MSE) of the users’ data via selective error-correction decoding. Finally, we derive a detailed system performance analysis that closely matches the obtained numerical results. We demonstrate that the presented system can more than double the number of active users, supported by the state-of-the-art systems. Large gains in terms of the minimal required signal-to-noise ratios (SNR)s are also demonstrated for a wide range of active user numbers.