Análisis multiobjetivo sobre DTNs

Abstract

En este trabajo se aborda el desafío de la comunicación espacial e interplanetaria a través del uso de redes tolerantes a demoras (DTNs) con un enfoque basado en procesos de decisión de Markov (MDPs). Se emplea la verificación de modelos probabilísticos para realizar un análisis multiobjetivo del modelo, con el objetivo de obtener planes de enrutamiento que minimicen el consumo de energía y el tiempo, al mismo tiempo que maximizan la probabilidad de entrega. Además, se trazan curvas de Pareto para examinar las diferentes optimizaciones posibles en el sistema y los compromisos inherentes entre los objetivos.

Delay tolerant networks (DTN) have been proposed and studied as an architecture fit for deep space and interplanetary networking. Based on orbital dynamics a contact plan can be known in advance to propose a competent scheduler. Of interest for this work are low-Earth orbit (LEO) constellations in which uncertainty in contact plans can arise from various sources; such as imprecise knowledge of the satellite positions and velocities, fault nodes and unpredictable changes in the atmospheric conditions. To solve routing in DTNs under uncertain contact plans Markov decision processes (MDP) have been studied and implemented to find schedulers with a good success delivery probability (SDP). This thesis analyzes further the modeling of DTNs with MDPs in order to include energy consumption and delivery time. Applying multi-objective analysis on MDPs using the STORM model checker we can expose underlying trade offs of specific DTNs. As for proof of concept a Lunar Road Ring architecture is analyzed in these three objectives.