A Little Prediction Goes a Long Way: Routing in Semi-Deterministic Delay Tolerant Networks

Realizing delay-capacity in intermittently connected mobile networks remains a largely open question, with state-of-the-art routing schemes typically focusing either on delay or on capacity. We show the feasibility of routing with both high goodput and desired delay constraints, with REAPER (for Reliable, Efficient, and Predictive Routing), a fully distributed convergecast routing framework that jointly optimizes both path length and path delay. A key idea for efficient instantiation of REAPER is to exploit predictability of mobility patterns, in terms of a semi-deterministic model which appropriately captures several vehicular and human inter-contact patterns. Packets are thus routed using paths that are jointly optimal at their time of arrival, in contrast to extant DTN protocols which use time-average metrics for routing. REAPER is also self-stabilizing to changes in the mobility pattern. A simulation-based evaluation confirms that, across the spectrum of ultra-light to heavy traffics, REAPER achieves up to 135% and 200% higher throughput and up to 250% and 1666% higher energy efficiency than state-of-the-art single-copy protocols MEED-DVR and PROPHET, which optimize a single metric only, specifically, expected delay and path probability respectively.