<p>The interest in delay tolerant networks (DTN’s) has grown considerably due to its various applications in the fields of military, critical situations, space, traffic, and many other areas. These networks often suffer from intermittent disruption and variable long delay due to factors such as mobility and energy. In DTNs there is no guarantee of end-to-end connectivity between source and destination, so the routing pattern in these networks is Store–Carry–Forward. In the literature, many multipath routing protocols are exploited to achieve the requirements; however, they suffer from low delivery rates and long delays. These constraints would degrade the performance of DTN’s services used in critical environments. In this paper to address this issue, the routing scheme based on reinforcement learning and PROPHET routing is proposed. The PROPHET protocol is one of the most popular delay tolerant networks protocols, and many methods have been proposed to improve this protocol. In the PROPHET protocol, there is a probabilistic criterion called the probability of delivery, which is calculated in each node A for each specific destination B. In this paper, a method for calculating this criterion using reinforcement learning is proposed. Given that the nodes have memory, and there are parameters such as the number of visits or delivery probabilities in the nodes, so these nodes can learn. As a result, the proposed scheme could achieve suitable delivery ratio and delay rate.</p>