Maximizing packet delivery ratio in stochastic routing: An analytical perspective

Stochastic routing offers a promising substitute to deterministic routing in wireless sensor networks, mainly due to its randomized nature. Still, most of the stochastic routing algorithms are heuristic, while the optimality for a given degree of randomness is difficult to find. In this paper, we propose a theoretical framework to find the optimal stochastic routing policy that maximizes the packet delivery ratio for a given degree of randomness in routing. Our model consists of directed acyclic networks with unreliable communication links. The degree of randomness is quantified by the global entropy of routing. By employing the Lagrangian approach, a simple iterative algorithm to find the optimal stochastic routing policy is proposed. This can be used as a benchmark for performance comparison of stochastic routing algorithms.