Optimization algorithms have made considerable advancements in solving complex problems with the ability to be applied to innumerable real-world problems. Nevertheless, they are passed through several challenges comprising of equilibrium between exploration and exploitation capabilities, and departure from local optimums. Portioning the population into several sub-populations is a robust technique to enhance the dispersion of the solution in the problem space. Consequently, the exploration would be increased, and the local optimums can be avoided. Furthermore, improving the exploration and exploitation capabilities is a way of increasing the authority of optimization algorithms that various researches have been considered, and numerous methods have been proposed. In this paper, a novel hybrid multi-population algorithm called HMPA is presented. First, a new portioning method is introduced to divide the population into several sub-populations. The sub-populations dynamically exchange solutions aiming at balancing the exploration and exploitation capabilities. Afterthought, artificial ecosystem-based optimization (AEO) and Harris Hawks optimization (HHO) algorithms are hybridized. Subsequently, levy-flight strategy, local search mechanism, quasi-oppositional learning, and chaos theory are utilized in a splendid way to maximize the efficiency of the HMPA. Next, HMPA is evaluated on fifty unimodal, multimodal, fix-dimension, shifted rotated, hybrid, and composite test functions. In addition, the results of HMPA is compared with similar state-of-the-art algorithms using five well-known statistical metrics, box plot, convergence rate, execution time, and Wilcoxon’s signed-rank test. Finally, the performance of the HMPA is investigated on seven constrained/unconstrained real-life engineering problems. The results demonstrate that the HMPA is outperformed the other competitor algorithms significantly.