Cross-layer optimizations for modeling and design of wireless networks
Rajive L. Bagrodia, Zhengrong Ji · 2006
In this dissertation, we exploit cross-layer optimization techniques to facilitate modeling and design of wireless networking systems. We proposed three solutions that respectively improve the modeling, design, and implementation aspects of wireless networking systems by exploiting cross-layer knowledge, particularly at the physical and MAC layers. First, we looked into the problem of improving the scalability of wireless network simulation without compromising its accuracy. Through theoretical analysis of physical and MAC operations, we quantified the effects of common optimizations such as propagation limit on the upper layer simulation results and identified parameters tolerable to most network studies. Further, we proposed optimization technique that utilizes the MAC layer states in each node to reduce the overhead of physical layer simulation. Our results showed that the proposed techniques reduced the simulation time of large wireless networks by up to 55 times. Next, we exploited the multi-rate support of wireless radios to improve wireless network capacity. We proposed a Medium Access Diversity (MAD) scheme that opportunistically exploits multiuser diversity via cooperation between the physical and MAC layers. Utilizing knowledge from both layers, we identified and addressed the challenges in the design and implementation of MAD's three phases: channel probing, data transmission, and receiver scheduling. Our results showed that MAD improved wireless LAN capacity by 30-120% over the best existing rate adaptation scheme, OAR. Finally, we utilized the characteristics of signal attenuation common to all wireless communications to design and implement a point-casting service (PCs) over existing wireless networks. The principle of PCs is to split and encode the downlink traffic and deliver them to, the client via a few surrounding access points (APs) to bound the service area within the intersection of communication coverage of the APs. Such intersection is further reduced via coordinated rate and power control at the physical and MAC layers of the APs. A prototype system is developed upon an 802.11a wireless LAN in a laboratory environment. Our results showed that PCs bounded service area to two meters near a client, and incurred average jitter of less than 1ms and less than 10% capacity loss. In summary, we addressed a broad set of problems through cross layer optimization, utilizing physical and MAC layer knowledge. Our experience clearly showed the benefits of such optimization in improving simulation scalability, exploiting multiuser diversity and providing novel service for wireless networks.