High Performance Medium Access Control Protocols for Decentralized Wireless Networks Using Local Coordination PDF Download

Author:
Publisher:
ISBN: 9781109249194
Category : Wireless communication systems
Languages : en
Pages :

Get Book

Book Description
Wireless networks differ from their wired counterparts in that communication between nodes takes place over a "link" using an RF, acoustic, optical, or other signal transmitted through the air or water instead of, as their name implies, a wire. This difference changes the frequency of transmission errors from extremely rare to almost constant, and introduces inter-node interference as a significant problem. Wireless networks are typically more limited than wired networks in terms of bandwidth, computational ability, power, and centralized management. Efficient handling of transmission errors and reducing interference are thus vital in maximizing network performance. This dissertation addresses two separate aspects of wireless networks with a common theme of low overhead, local coordination between nodes, and often using inferences or even informed guesses to make decisions. To address the problem of transmission errors, we study two medium access control (MAC) protocols that use minimal-overhead, local coordination schemes to allow cooperation between neighboring nodes: one with and one without a cooperation-enabled physical layer. To address the problem of interference, we study two closely related MAC protocols that use local coordination between neighboring nodes to build an interference-free transmission schedule, for (1) supporting latency-sensitive applications over long routes in mesh networks, and (2) increasing channel utilization and energy efficiency in underwater acoustic networks. Our first work focuses on mobile ad hoc networks where if any link in a route fails, multiple fruitless attempts are currently made by most of the existing MAC protocols to use the failed link before reporting failure to the routing layer and/or attempting local recovery. The high frequency of link errors between mobile nodes requires rapid recovery to provide acceptable performance. We design CIFLER, a cross-layer approach which uses enhanced channel reservation messages to allow alternate nodes to immediately elect themselves using only inferred neighbor information. This self-election avoids reliance on individual links, and uses diversity to minimize the impact frequent link errors have on delay, energy efficiency, and the functioning of upper layer protocols. We show via both analysis and simulation that CIFLER provides better performance in typical MANET scenarios. Unlike other local recovery schemes, CIFLER uses only a minor modification to IEEE 802.11 DCF, does not suffer from duplicated messages, allows neighboring nodes to almost immediately learn the information needed to assist in the recovery of existing routes, and does not require additional hardware, delays, or control messages. Our second work applies the same concept of inferred neighbor information to cooperative communications, where the signals of simultaneous transmissions by multiple nodes constructively combine in the wireless medium. Studies on the physical layer capabilities (via either information theory or numerical analysis) have shown the significant performance improvements of cooperative communications. However, these studies ignore both the overheads incurred in real implementations of the cooperative techniques at the physical layer and their interactions with higher layer protocols in a networking context. We implement a path-centric MAC protocol that uses minimal control messages to reserve a multi-hop path between source and destination nodes, and perform coordination between relay nodes. We then realistically study the performance of cooperation in networking scenarios by taking into account overheads incurred at the physical, MAC, and network layers. Simulations demonstrate that significant performance improvement can be achieved by employing cooperation. We also demonstrate the overheads which challenge the effectiveness of such schemes in real networks. Our third work deals with the issue of interference and transmission scheduling in mesh networks, where links are generally reliable if no interference is present. In current wireless networks, access to the shared wireless medium is controlled via either a TDMA- or a CSMA-based scheme. While usable in single-hop networks, these techniques are often far from optimal, and result in significant per-hop and per-packet delay and jitter, making multi-hop wireless mesh networks a particularly harsh environment for real-time, isochronous applications such as VoIP. We present a new time-based MAC protocol, FLASHR, for wireless mesh networks carrying delay-sensitive isochronous traffic. In our scheme, nodes use simple local coordination mechanisms to form adaptive transmission schedules which attain the desired quality of service. Simulations show that our scheme achieves near-optimal capacity, minimal jitter, and a weaker correlation between route length and end-to-end delay. Our final work adapts the FLASHR MAC protocol for use in underwater acoustic networks. A time-based MAC has potential advantages over FDMA and CDMA approaches in terms of hardware simplicity, energy efficiency, and delay. Unfortunately, the channel utilization of existing TDMA and CSMA acoustic MAC protocols is generally low due to the long propagation delays of acoustic signals. We argue that several ideas taken from RF protocols, including exclusive channel access, are either unnecessary in acoustic networks or must be redefined. We design UW-FLASHR, a modification to FLASHR which uses additional local control messages to create a time-based MAC protocol for acoustic networks which does not require centralized control, tight clock synchronization, or accurate propagation delay estimation. Our results show that UWFLASHR achieves higher channel utilization than the maximum utilization possible with existing time-based exclusive-access MAC protocols, particularly when the ratio of propagation delay to transmission delay is high, or data payloads are small.