Evolutionary Multitask Optimization for Coupled Continuous Decision Spaces: A GPU-Accelerated Pre-optimization for Movable-Antenna-and-RIS-aided Systems

Next-generation 6G wireless systems aim to improve spectral efficiency and user fairness by leveraging Movable Arrays (MA) and Reconfigurable Intelligent Surfaces (RIS). However, the joint optimization of antenna positions and RIS phase shifts leads to high-dimensional, non-convex problems that are computationally prohibitive for real-time deployment within channel coherence times. This paper proposes a pre-optimization method framework using an adaptive Multifactorial Evolutionary Algorithm (MFEA). We consider multiple propagation conditions and formulate three concurrent objective-driven tasks that target sum-rate maximization, user fairness (min-rate maximization), and energy efficiency, respectively. By exploiting inter-task knowledge transfer, the proposed adaptive MFEA evolves a single antenna layout that is robust across diverse operating conditions. We incorporate an adaptive Random Mating Probability (RMP) mechanism based on transfer-offspring survival to regulate cross-task knowledge transfer. Experimental results show that the pre-optimized layout recovers over 98% of the performance of real-time optimization while effectively eliminating online optimization latency. Compared with conventional uniform square planar arrays, our approach yields consistently stronger performance, providing a robust hardware configuration for dynamic 6G deployments.