Hierarchical structure-based joint operations algorithm for global optimization

Joint operations algorithm (JOA) is a metaheuristic algorithm based on joint operations strategy in military theory, which incorporates three core operations - offensive, defensive and regroup - and has excellent performance in global optimization problems. To enhance the optimization performance of the original JOA, we re-examine the positioning of the three core operations in balancing global exploration and local exploitation, and propose a hierarchical structure-based JOA variant (abbreviated as HSJOA) by adjusting their execution mechanism. In addition, we redesign the three core operations to give full play to their synergy effect. In the new offensive operations, we simplify its specific execution strategy, but introduce an adjustment parameter to retain valuable position information. In the modified defensive operations, we integrate the influence of officer and elite comrades on soldiers to design a Gaussian distribution-based conservative defensive strategy and one Cauchy distribution-based aggressive defensive strategy. In the novel regroup operations, we replace the original random division strategy with a sorting-based division scheme. To evaluate the optimization performance of HSJOA, we conduct comparison experiments using nine excellent algorithms to deal with four real-life optimization problems and thirty test functions from IEEE CEC 2014 testbed. The comparison results show that the overall optimization performance of HSJOA is significantly better than the original JOA, the outstanding variants (TAPSO and GODE) of two well-known algorithms and four recently proposed algorithms (EO, SOA, MPA and WOA), while inferior to two winners of IEEE CEC competition (L-SHADE and EBOwithCMAR), but HSJOA clearly outperforms L-SHADE and EBOwithCMAR in terms of the runtime consumption.

Automated summary

Joint operations algorithm (JOA) is a metaheuristic algorithm that utilizes offensive, defensive, and regroup operations to optimize global problems. In order to improve its performance, a hierarchical structure-based variant called HSJOA is proposed by adjusting the execution mechanism of the core operations and redesigning their strategies. Experimental results on real-life optimization problems and test functions demonstrate that HSJOA outperforms both the original JOA and other algorithms, achieving better optimization performance and runtime consumption than L-SHADE and EBOwithCMAR.