In computer simulations of next-generation wireless communications, engineering researchers have demonstrated that modifying a standard genetic algorithm with mechanisms derived from ant colony optimization significantly improves the configuration of reconfigurable intelligent surfaces in multiple-input multiple-output systems. The study, conducted by Laith Ibrahim, Mohd Nazri Mahmud, Mohd Fadzli Mohd Salleh, and Ashraf Al-Rimawi, addresses the complex computational challenge of adjusting numerous reflective elements on smart radio surfaces to maximize received signal strength. By embedding pheromone-guided selection and weighted recombination directly into the genetic algorithm’s crossover stage, the hybrid approach achieved steady-state channel gains up to 2.4 decibels higher than alternative algorithms while cutting running times by more than half.
Reconfigurable intelligent surfaces, often abbreviated as RISs, are engineered planar arrays composed of numerous passive reflecting units designed to assist future sixth-generation wireless networks. In environments where the direct propagation path between a multi-antenna transmitter and a stationary receiver is severely obstructed, these surfaces can intercept incoming radio waves and steer them toward the receiver by introducing controlled phase shifts across each element. However, identifying the optimal combination of phase adjustments across hundreds of reflecting units represents a non-convex, high-dimensional optimization challenge. The difficulty is magnified when configuring discrete phase levels, which transform the optimization into a mixed-integer nonlinear programming problem characterized by coupled spatial variables and tight hardware constraints.
To navigate these challenging search spaces, engineers frequently deploy meta-heuristic algorithms such as the classical genetic algorithm and particle swarm optimization. Genetic algorithms model natural evolutionary processes through iterative cycles of candidate selection, genetic crossover, and mutation, offering robust global exploration across high-dimensional parameter spaces. Despite their broad exploratory capabilities, standard genetic algorithms exhibit notable deficiencies when applied to phase shift tuning, including sluggish convergence rates, high sensitivity to internal control parameters, and poor exploitation of promising local solutions. Standard crossover operators rely on uniform or random genetic exchanges between two parent candidates, lacking an internal memory mechanism to preserve and exploit high-performing configuration traits identified during previous generations.
The research team addressed these operational bottlenecks by introducing a lightweight hybridization strategy that integrates the principles of ant colony optimization strictly into the crossover operator of a genetic algorithm. In nature, ant colonies locate efficient foraging routes through the deposition of pheromone trails, creating an indirect communication system where advantageous pathways receive progressive reinforcement while less productive paths decay through natural evaporation. Translating this behavior into algorithmic design, the proposed method evaluates parent candidates within a mating pool using both objective fitness values and dynamic pheromone concentrations. Rather than executing two separate algorithms sequentially or maintaining parallel populations, this targeted architecture retains the overall evolutionary structure of the genetic algorithm while using historical search data to systematically direct the formation of offspring.
Under the enhanced crossover mechanism, parent individuals are chosen probabilistically according to a decision rule that balances their accumulated pheromone intensity and normalized heuristic fitness information. The algorithm then selects more than two parent candidates and synthesizes new candidate solutions through a multi-parent weighted average recombination scheme. In this step, phase shift genes contributed by superior parents carry greater numerical weight in determining the final offspring values, followed by a modulo operation to ensure the resultant phases remain strictly bounded within the physical angular range of zero to two pi radians. Subsequent fitness evaluations of the generated offspring determine the magnitude of new pheromone reinforcement, while a defined evaporation parameter prevents excessive pheromone concentration on local optima.
To evaluate the internal mechanics of the search process, the authors established an analytical framework based on Shannon entropy to quantify population diversity across successive iterations. By dividing the phase domain into discrete genotypic bins and tracking the probability distribution of candidate phases, the framework models entropy decay as a first-order differential equation approaching a steady-state lower bound. This formulation yielded a normalized balance metric tracking the transition from initial wide-ranging exploration to localized exploitation. Numerical assessments revealed that the enhanced genetic algorithm attained a 95 percent transition toward exploitation in approximately 6 iterations across both small-scale and large-scale networks, whereas the standard genetic algorithm required 66 to 83 iterations and particle swarm optimization required 40 to 43 iterations.
The performance of the hybrid technique was benchmarked against the standard genetic algorithm and particle swarm optimization across two indoor simulated wireless topologies modeled under Rician fading channels with dominant line-of-sight components. The small-scale configuration comprised a 9-antenna transmitter, a 100-element intelligent surface, and a 4-antenna receiver, while the large-scale configuration incorporated a 64-antenna transmitter, a 225-element surface, and a 9-antenna receiver. The evaluation investigated both continuous phase adjustments and discrete phase quantization levels fixed at eight states. Convergence was defined as the stage where the performance difference between two consecutive iterations dropped below 0.01 decibels, with execution runtimes empirically recorded on a shared computing platform utilizing general-purpose desktop hardware.
Across all evaluated scenarios, the enhanced algorithm converged consistently within 7 iterations. In the continuous phase shift setup for the large-scale network, it converged roughly 12.3 times faster than the original genetic algorithm, which took 86 iterations, and 6.3 times faster than particle swarm optimization, which required 44 iterations. This rapid stabilization translated into significant gains in signal quality, yielding a steady-state channel gain 1.3 decibels above the standard genetic algorithm and 2.4 decibels above particle swarm optimization. In the discrete phase setting, the enhanced algorithm provided channel gain advantages of 1.0 decibel over the baseline genetic algorithm and 1.8 decibels over particle swarm optimization, demonstrating that the structural advantages persist when phase shifts are restricted to practical, quantized hardware states.
The empirical runtime evaluations showed corresponding improvements in computational efficiency despite the extra calculations required for pheromone updates and multi-parent gene averaging. In the large-scale continuous configuration, the hybrid algorithm reached the convergence threshold in 0.20 seconds, representing a 74 percent reduction compared to the 0.78 seconds required by the baseline genetic algorithm and a 52 percent reduction relative to the 0.42 seconds measured for particle swarm optimization. In the small-scale system, the enhanced optimizer achieved convergence in 0.055 seconds, outperforming the standard genetic algorithm at 0.14 seconds and particle swarm optimization at 0.095 seconds. The authors noted that while these runtimes reflect software execution on general-purpose processors, specialized FPGA or hybrid hardware implementations could potentially compress processing latency into the 1 to 10 millisecond range.
Scalability assessments highlighted an expanding performance margin as network dimensions increased. Transitioning from the small-scale to the large-scale layout widened the hybrid method’s channel gain advantage over the original genetic algorithm from 0.7 decibels to 1.3 decibels in continuous mode, and from 0.4 decibels to 1.0 decibels in discrete mode. Sensitivity analyses further established that the enhanced algorithm maintained rapid convergence and high channel gains when the population size was reduced to 50 individuals, whereas the baseline genetic algorithm required a population of 200 individuals to approach similar channel gain levels, albeit at the expense of prolonged iteration counts. The hybrid algorithm also displayed minimal sensitivity to variations in the mutation rate, avoiding the instability observed in the standard genetic algorithm under elevated mutation settings.
To evaluate operational resilience against physical channel impairments, the researchers introduced complex Gaussian estimation errors into both the transmitter-to-surface and surface-to-receiver channel matrices. When evaluated under imperfect channel state information, the enhanced algorithm experienced only modest reductions in final channel gain and maintained steep early convergence trajectories, whereas the baseline algorithms suffered pronounced performance degradation and erratic fluctuations. The authors concluded that while the simulation results establish strong computational and scalability benefits for static, line-of-sight environments, future evaluations must address mobility-induced channel aging, non-line-of-sight scattering conditions, nonlinear amplitude-phase coupling in physical hardware diodes, and experimental validation on over-the-air programmable surface testbeds.
Subject of Research: Phase shift optimization in reconfigurable intelligent surface-assisted MIMO wireless networks
Article Title: An enhanced genetic algorithm with ant colony optimization for phase shift design in RIS-assisted MIMO networks
Article References: Ibrahim, L., Mahmud, M. N., Salleh, M. F. M., & Al-Rimawi, A. (2026). An enhanced genetic algorithm with ant colony optimization for phase shift design in RIS-assisted MIMO networks. Journal of King Saud University – Engineering Sciences, 38(4), Article 29. https://doi.org/10.1007/s44444-026-00102-5
Image Credits: AI Generated
DOI: 10.1007/s44444-026-00102-5
Keywords: Reconfigurable Intelligent Surfaces, Genetic Algorithm, Ant Colony Optimization, MIMO Networks, Wireless Communications, Phase Shift Optimization, Particle Swarm Optimization, Channel Gain, Meta-Heuristic Optimization, Exploration-Exploitation Balance, Shannon Entropy, 6G Wireless Systems, Computational Efficiency, Beamforming, Discrete Phase Shifts, Channel Estimation Errors
Cite Scienmag News
Juliet Wilcox. (November 21, 2026). Hybrid Genetic Algorithm Speeds Phase Shift Optimization for Smart Radio Surfaces. Scienmag. https://scienmag.com/hybrid-genetic-algorithm-speeds-phase-shift-optimization-for-smart-radio-surfaces/
Juliet Wilcox. "Hybrid Genetic Algorithm Speeds Phase Shift Optimization for Smart Radio Surfaces." Scienmag, 21 November 2026, https://scienmag.com/hybrid-genetic-algorithm-speeds-phase-shift-optimization-for-smart-radio-surfaces/. Accessed 21 November 2026.
Juliet Wilcox. "Hybrid Genetic Algorithm Speeds Phase Shift Optimization for Smart Radio Surfaces." Scienmag. November 21, 2026. https://scienmag.com/hybrid-genetic-algorithm-speeds-phase-shift-optimization-for-smart-radio-surfaces/








