An optimization of neighbor discovery in mobile ad hoc networks

This thesis aims to improve the performance of Mobile Ad Hoc Networks (MANETs) that use soft-state signaling for neighbor discovery, specifically OLSR. In particular, it uses simulations and previous work based on experiments conducted on a physical test bed to study the behavior of the neighbor-discovery algorithm used by OLSR to identify and explore ways to optimize the neighbor-discovery process. Candidate optimizations include the actual and relative settings of the refresh and expiry timers used by the algorithm. Previous studies focused on understanding how the settings of the algorithms Hello Interval (δ) and TC Interval refresh timers affect network performance in terms of protocol message overhead and network throughput. In contrast, this thesis investigates how the setting of the Hello Validity (τ ) timer relative to the setting of the Hello Interval timer affects network performance in terms of overall network packet loss and how the setting of the Hello Interval timer affects energy consumption vs. packet loss. Results obtained from our simulations indicate that the relationship between the settings of these timers impact network performance in terms of the percentage of overall packet loss. In particular, we discovered that: (1) For Hello Validity values smaller than 2δ, as the Hello Validity approaches the value of the Hello Interval, the percentage of packet loss increases due to the proximity of these parameters and collisions; and (2) setting the Hello Validity timer to twice the value of the Hello Interval timer results in a configuration with little packet loss. We attribute these results to a phenomenon that we call neighbor flapping , where neighbor information for a node that is within range is repeatedly placed in and then evicted from the neighbor sets of other nodes. In terms of energy consumption, we discovered that as the Hello Interval timer increases, energy consumption decreases but packet loss significantly increases