<p>Rapid item distribution to applicants, increasing decision-making speed, improving decision-making quality, and aiding in planning and prioritizing central distribution point establishment. During severe crises like earthquakes, decisions on locating, allocating, and distributing vital items are crucial for humanitarian relief managers. Items are dispatched to demand areas via central distribution points. Due to limited resources during crises, all central distribution points cannot be set up simultaneously. Hence, prioritizing their establishment is essential. Setting up these points significantly enhances response time and service quality. Agile relief systems are pivotal in improving service quality. Without decision-making tools, comprehensive planning for such issues is challenging. Hence, this study aims to provide a reinforcement learning-based hybrid decision support system for humanitarian relief chains during the crisis to support us in decision-making based on environmental conditions. In this instance, there will be a boost in the velocity and efficiency of the decision-making process. The Q-learning method is the core of processing and computations. The Q-learning method was compared with a random walk, ε -greedy, and simulated annealing algorithms for a simulated problem with high iterations. The comparison results indicate that the algorithm under analysis provides proper efficiency in the routing process. This algorithm was used as the major core of the study to design a hybrid decision support system for earthquake crises.</p>