The evolution of 6G has brought distributed multi-input multi-output (D-MIMO) near-field communications into the spotlight, owing to their potential to significantly enhance wireless capacity and spectral efficiency. This study delves into exploring the optimal characteristics of distributed antenna deployment in D-MIMO systems through the optimization of access point (AP) placement and subarray topology. To guarantee the stability of system configuration, we derive an approximate ergodic sum rate with high accuracy under near-field spherical wavefront propagation that serves as the optimization criterion. Confronting the non-convex challenges, we adopt an approach integrating successive convex approximation (SCA), gradient descent, and alternating optimization methods, which enables us to attain a near-optimal solution with low complexity. Numerical simulations highlight the significant superiority of our proposal to traditional baseline approaches, revealing key characteristics and providing crucial insights for AP deployment and array layout optimization. In densely clustered user scenarios, APs should be deployed near the user aggregation centroid, which may not align with the geometric center. For dispersed user distributions, AP deployment positions should be adjusted more flexibly to regional statistical characteristics rather than confined to central points to accommodate specific user distribution features. Moreover, the array topology exhibits a pattern of alternating dense and sparse distribution, diverging from the attributes observed in centralized MIMO (C-MIMO) architectures, where a consistent medium density is flanked by regions of sparsity.

Quantized Convolutional Neural Network (QCNN) is an attractive approach that reduces hardware overheads, especially for energy-constrained systems. However, existing QCNNs still require non-trivial hardware resources and memory capacity in order not to compromise model accuracy. To address this issue, we propose an antiferromagnetic magnetic random-access memory (ARAM)-based processing-in-memory (PIM) system, leveraging bit-level sparsity. Three optimization techniques are proposed to optimize hardware resource utilization while preserving CNN accuracy. Firstly, the ARAM-based memory subsystem allows dynamic adaptation of variable bit-width across CNN layers. Secondly, the bit-level accelerator employs the bit-fusion format engineered for processing data from the ARAM subsystem. Thirdly, a customized data path within the RISC-V core guarantees efficient instruction processing to the ARAM-based memory subsystem and bit-level accelerator, enabling optimal bit-level data transmission and computation. Experimental results demonstrate that this design remarkably reduces data movement by 50%-83% across existing CNNs. Compared to state-of-the-art designs, it enhances throughput and latency by an average of 5x and 10x, respectively. In addition, this design achieves speedups between 1.63x and 2.96x, outstripping other designs in AlexNet, VGG16, and ResNet18 benchmarks.

Using the polarization as a scale, we compensated for the azimuth scale effect on the building identification in urban areas. With the increase of polarization dimension (from a single polarization, dual-polarization, to polarimetric data with the rotation of azimuth angle) or change of polarization scale, the azimuth scale effect was gradually resolved.

A normalized function was proposed as the framework to study the heterogeneity of urban radar targets. The function consisted of three types of scales, the physical size of a target, heterogeneity in dielectric constant, and target geometry. The azimuth direction or azimuth scale, one component of the target geometry was exampled to investigate the influence of azimuth orientations of buildings on radar backscattering. The results should advance SAR application and theory.