Distributed Metaheuristic Portfolio Selection for Adaptive Routing in Delay/Disruption-Tolerant Networks

Abstract

Delay and disruption tolerant networks arise in environments where connectivity is intermittent, mobility is high, and end-to-end paths rarely exist for a sufficient duration. In such settings, routing decisions depend on stochastic contact patterns, constrained buffers, and heterogeneous node behaviors. Traditional routing schemes either rely on epidemic replication, which increases resource consumption, or exploit single carefully designed heuristics, which may be fragile under non-stationary conditions. At the same time, metaheuristic optimization techniques offer flexible search procedures that can adapt routing choices to empirical performance signals but tend to focus on single algorithms and centralized control. This paper considers a distributed portfolio view of metaheuristics for routing in delay and disruption tolerant networks, where each node maintains and updates a mixture of routing metaheuristics that jointly drive forwarding decisions. The study explores how portfolio weights can be learned online from local contact histories and performance indicators, while accounting for limited information, communication costs, and heterogeneous network regions. The proposed framework models routing as a linear optimization problem that couples contact opportunities with portfolio allocation decisions and uses distributed feedback to modulate exploration and exploitation among candidate heuristics. The paper discusses algorithmic design, convergence properties at a qualitative level, and expected trade-offs in delivery ratio, latency, and overhead. Analytical modeling and conceptual evaluation highlight how a distributed metaheuristic portfolio can adjust to changing mobility patterns without assuming global knowledge or centralized coordination. The discussion emphasizes conditions under which such portfolios may offer balanced behavior across diverse delay and disruption tolerant network scenarios.