Adaptive DEBKA for joint beamforming and spectrum sharing in 6G massive MIMO IoT networks

Abstract

Joint hybrid beamforming and underlay spectrum sharing in sixth-generation (6G) massive multiple-input multiple-output (MIMO) Internet-of-Things (IoT) networks yields a non-convex mixed continuous-discrete problem involving constant-modulus analog precoding, subchannel-dependent digital precoding, and binary link activation. This paper develops DEBKA, a training-free optimizer that assigns different search roles to two complementary mechanisms: a BKA-inspired Cauchy-Gaussian attack for local refinement and adaptive DE/current-to-pbest migration for population-difference exploration. These operators are coupled with success-history parameter updates, a rejected-parent archive, constant-modulus-preserving decoding, power normalization, and feasibility-first selection. The methodological contribution therefore lies in the phase-specific BKA-DE architecture and its problem-specific constraint pipeline, rather than in claiming the individual invention of JADE components. Simulations use a simplified geometric mmWave channel model inspired by 3GPP TR 38.901 and compare DEBKA with ten locally implemented baselines over 30 independent runs. Under the default configuration, DEBKA attains 15.87 ± 0.19 bps/Hz, 27.5% above MO-AltMin and 8.1% above BKAPI, and reaches the 1% convergence criterion in 185 ± 12 iterations. Ablation, sensitivity, robustness, and runtime results support the contribution of the coupled search and constraint-handling design within the evaluated scenarios. These results do not establish global optimality or deployment readiness; validation with imperfect channel state information, standardized channel generation, and hardware experiments remains necessary.