Comparing OLSRd, d2 and -Q Using an Emulated 802.11 or TDMA MANET in a 2-ring Topology

摘要

A Q-learning solution called OLSR-Q was proposed and tuned by McAuley et al. in 2012 to improve the performance of Optimized Link State Routing (OLSR) using a TDMA MANET simulation and a testbed with real WiFi cards. Here, the use of either IEEE802.11 or TDMA radios are compared using real time emulations based on the open source EMANE platform implemented on a cluster of virtual machines. The open source OLSRd and newer OLSRd2 are analyzed and compared to our implementation called OLSRd2-Q2 of the OLSR-Q algorithm. The benefit of our OLSRd2-Q2 is clear when using TDMA radios but not when using IEEE802.11 radios. These preliminary results confirm the potential and the practical challenges of cognitive approaches to improve the MANET performance of the open source implementation OLSRd2.