Multi-Objective Cognitive Routing in Space DTNs

In space networking and other challenging environments, optimizing bundle routing is crucial. Traditional routing objectives focus on finding the shortest path or achieving the earliest delivery time. However, with the growth of these networks in size and diversity of applications, there is a need to expand the scope of routing objectives. In this paper, we explore a decentralized bundle routing method with multiple objectives. These objectives include bundle response time, loss ratio, and a general cost that reflects the monetary cost rate associated with the utilization of network links. We also consider that certain links may be offered as on-demand or reserved services. To address these challenges, we propose a combined routing objective and study its performance. We then apply this objective to determine learning rewards for a reinforcement learning agent. Our results, based on a proof-of-concept using a standard Delay- Tolerant Networking (DTN) implementation, demonstrate the effectiveness of the proposed method. By accommodating multiple objectives, the method offers a flexible and adaptive solution for online routing optimization.