<p>Four improvement strategies are proposed in this paper to address the shortcomings of artificial rabbit optimization, which suffers from weak exploitation performance and tends to stagnate in the later stages. At first, a good point set strategy is employed to generate the initial population for the purpose of improving its initial solution quality and global exploration capability. In the exploitation phase of ARO, roulette cave selection strategy is proposed, which enables more dominant individuals to be selected and guarantees that more local exploitation is carried out in the exploitation phase. At the same time, the rest of the individuals can also be selected, which provides global exploration capability and achieves a balance between global and local search. The worst escaping strategy is employed to further refine the quality of poorer individuals. By absorbing the knowledge of other individuals, the worse individual realizes the restructuring of its own position, which then strengthens the population diversity and accelerates the convergence of the algorithm. Furthermore, we provide a population-based differential restart strategy in order to offer the population a chance to escape from the localization trap. The performance of MEARO is evaluated using test functions from the CEC 2017 and CEC 2022 test sets. The experimental results show that MEARO has higher robustness and superiority than RIME, PSA, SMA, COA, DTSMA, AFDBARO, FDBARO, LCAHA and basic ARO. Lastly, MEARO outperforms its competitors in the UAV path planning problem, showing its great potential in practical applications.</p>