Network Selection in Device-to-Device Communication Using Genetic AI

D2D communication has emerged as one of the
key enablers for 5G and Beyond networks in ultra-dense
deployments, proximity services, and low-latency applications.
Selection of an appropriate communication mode and access
network remains however a challenging multi-criteria optimization problem for D2D pairs in heterogeneous environments. This
paper introduces a Genetic AI framework for network selection
in D2D communication, which employs a Genetic Algorithm
(GA) as the optimization core. The proposed framework will
jointly consider user Quality-of-Service, energy consumption,
interference to cellular users, and mobility-induced instability.
First, the problem is formulated as a constrained multi-objective
optimization; thereafter, an efficient chromosome encoding for
D2D mode and network interface selection will be presented.
It will further integrate the adaptive crossover and mutation
with heuristic seeding and local search. The proposed method
is compared with classical MADM schemes, such as SAW,
TOPSIS, and a machine-learning baseline by gradient boosting,
in evaluations. Simulation results in a 5G-like heterogeneous cell
layout show that the proposed Genetic AI approach improves
the spectral efficiency by up to 15%, reduces the handover
rate by approximately 25–30%, and achieves superior energythroughput trade-offs compared to reference schemes. Further,
this paper discusses the complexity, signalling overhead, and how
the proposed framework can be extended to federated and digitaltwin-assisted operation for 6G scenarios.