A Monte Carlo hyper-heuristic algorithm with low-level heuristics reward prediction for missile path planning

Missile path planning under multiple aircraft relay guidance is important in long-range air-to-ground strikes. Traditional meta-heuristic algorithms applied in path planning problems lack flexibility in the algorithm iteration process, and current hyper-heuristic (HH) algorithms have difficulty estimating the performance of low-level heuristics (LLHs) applied to the population in different states. This study proposes a Monte Carlo hyper-heuristic (MCHH) algorithm, which is adaptive to various path planning scenarios. The LLH set contains 18 LLHs generated from the basic operators in three meta-heuristic algorithms. The high-level strategy (HLS) evaluates the states of individuals and the reward of the LLH applied to each individual. A discrete state-action-reward table is used to predict the effectiveness of different LLHs and thus determine the optimal LLH applied in iterations. The table is trained through the MC method. The results of simulation cases and algorithm comparison demonstrate the efficiency and superiority of the MCHH algorithm.