Online Algorithms for Adaptive Optimization in Heterogeneous Delay Tolerant Networks

We study in this paper heterogeneous delay tolerant networks formed by different classes of nodes. We assume that two hop is the forwarding strategy to deliver messages from a source node to destination node and the goal is to optimize the probability of delivering a message in the presence of different classes of mobiles, while satisfying a given energy budget. Using our model, in particular, we transform the joint energy constraint into separate constraints, one per class. This allows us to characterize the optimal strategies and provides us with a suitable framework for the design of multi-dimensional stochastic control algorithms that achieve optimal performance at runtime in spite of the lack of full information on the network state and in presence of different classes of mobiles. A thorough analysis of the convergence properties and stability of our algorithms is presented.