One of the promising technologies for sending high data rates over long distances with compact antennas and low power usage is optical free-space communications (OFSC). The high-speed links between satellites, up-and down-links from and to ground stations, and other links between aeronautical vehicles or in deep space communications (DSC) are all potential scenarios for future applications. Various effects severely degrade the signal and capabilities of OFSC, such as laser beam spreading, wandering and scintillation. The first part of our study discusses the emulation concepts of wandering impact. Then, we use the PILab numerical simulation platform to visualize the effective parameters and simulate atmospheric optical propagation especially for wandering effect. Noteworthy, machine learning (ML) algorithms have recently been considered promising approaches for solving highorder nonlinear equations and comparing the activation function for high-order equations. We propose a laboratory testbed with a novel wavefront aberration compensation technique using Machine Learning, especially ML have a good chance to support developing and testing processes related to adaptive nonlinear characterization for OFSC systems. The turbulence generator's development, the testbed's concept, and its optical design are illustrated. The conceptual principles of the testbed make it possible to reproduce test scenarios with realistic ratios of aperture size over the Fried parameter.