Emperor penguin optimizer (EPO) is a population-based meta-heuristic approach that is inspired by the huddling behavior of emperor penguins. EPO suffers from the premature convergence problem, in which it gets trapped in local optima that prevents it from reaching the global optimum safely. The initial population generation directly affects diversity and convergence during the algorithm's execution in the multidimensional search space. The standard EPO uses a traditional randomization approach to generate the initial population, which cannot cover all regions of the search space. However, a low-discrepancy sequence such as Sobol ensures coverage of all potential regions in the search space and distributes the initial positions more evenly than traditional randomization. Moreover, minimizing the maximum SLL significantly enhances the reliability of the communication system by reducing the received noise and interference. This can be achieved by controlling the excitation currents (i.e., amplitude and phase) of the antenna elements while maintaining a fixed spacing between them in the array. This paper proposes the Sobol-based EPO (S-EPO) and employs the Rectangular Antenna Array (RAA) Sidelobe Level (SLL) optimization problem as a case study to verify its effectiveness. In this study, the excitation amplitudes of different RAAs are optimized to minimize the maximum SLL. The results from the various simulations show that S-EPO performs better in reducing the maximum SLL and improving the convergence rate during the optimization process. Here, the stability of S-EPO is compared with that of standard EPO and five benchmark algorithms: Aquila Optimizer (AO), COOT Optimization Algorithm (COAO), Jaya Algorithm (JA), Sine Cosine Algorithm (SCA), and Salp Swarm Algorithm (SSA). The statistical analysis shows that S-EPO achieves better performance among all compared algorithms.