Hybrid Hunger Games Search optimization using a neural networks approach applied to UAVs

Optimization methods like population-based algorithms are valuable when applied to multidimensional and nonlinear problems. Many engineering problems, such as controller parameterization, can be addressed using population-based algorithms since these parameters are usually found through essays, resulting in high time and resource consumption. Population-based algorithms need to define the range within which the search for the best solution is performed, known as the search space. However, due to the nonlinear nature of the systems to which these controllers are applied, there is no certainty about the search space that must be defined. This study proposes a hybrid optimization strategy that couples the Hunger Games Search (HGS) metaheuristic with an unsupervised Self Organizing Map, Kohonen Neural Network, to improve trajectory-tracking control of unmanned aerial vehicles (UAVs) transporting cable suspended loads. In the proposed NNHGS, the HGS algorithm seeks the controller gains that minimize Root Mean Square tracking Error (RMSE). At the same time, the neural network continuously reshapes the search intervals according to the evolving tracking performance. By expanding the exploration into parameter regions beyond the initial bounds, the NNHGS finds high-quality solutions that standard HGS excludes. The simulation results obtained with a Super Twisting Sliding Mode Controller (STSMC) show a reduction in the final tracking error from RMSE=0.0480 with HGS to RMSE = 0.0204 by NNHGS, along with enhanced disturbance rejection and rapid adaptation to parameter changes. These gains highlight the suitability of this method for real-world missions such as logistics, disaster relief, or remote inspection, where UAVs must remain stable under uncertain or parameter-varying conditions.