A Low-Complexity Noncoherent Uplink for IoT Devices in Massive MIMO Systems

This paper presents a noncoherent massive multiple-input multiple-output (MIMO) uplink scheme designed for low-complexity Internet of Things (IoT) devices. We consider a scenario where a single device equipped with two antennas transmits short packets to a base station with a large antenna array. To minimize latency and avoid pilot overhead, we introduce a structured space-time constellation design. Our key innovation is a parametric constellation framework that jointly encodes information in amplitude, mixing angle, and phase within an Alamouti structure. By optimizing the Kullback-Leibler (KL) divergence, we formulate the design as a tractable problem of selecting parameters for geometric and arithmetic sequences, rather than a complex search over all possible constellations. This results in a scheme specified by a few parameters, enabling both minimal storage and a low-complexity sequential maximum likelihood detector. Theoretical and numerical results demonstrate that our design achieves a larger minimum KL divergence and a superior symbol error rate compared to benchmarks, including Riemannian distance-based codes, KL-optimized single-input multiple-output schemes, and noncoherent pulse amplitude modulation, particularly in the high-SNR regime, where it eliminates error floors.