| This dissertation proposes new principles for designing and performance evaluation for spread spectrum based ad hoc networks. We first highlight the advantages of spread spectrum, in the form of Code Division Multiple Access (CDMA), in handling quality of service (QoS) requirements, enhancing energy efficiency, and enabling spatial multiplexing of bursty traffic. Then, based on stochastic geometric models and simulation, we show the ALOHA-like random channel access and 802.11-like simple contention and handshaking based schemes are poor at achieving good capacity or efficient spatial reuse, especially under bursty and heavy load. We show that this is because the closest interferers severely penalize the performance of the network, particularly for a direct sequence CDMA (DS-CDMA) system. Therefore, it is necessary to reconsider system design for spread spectrum ad hoc networks. To this end, we consider improving system performance at different network layers. At the physical layer, we first propose to use interference cancellation techniques, in particular, successive interference cancellation (SIC), at receivers to handle strong nearby interferers. Our analysis not only shows the significant improvement on capacity from SIC but also indicates that just canceling a few nearest interferers will provide most of the performance gain. Therefore, SIC is particularly suitable for DS-CDMA ad hoc networks to enhance capacity, incurring only a small amount of extra complexity. In addition, at the MAC layer, we show how idealized contention resolution among randomly distributed nodes results in clustering of successful transmitters and receivers, in turn leading to efficient spatial reuse. This motivates explicitly inducing clustering among contending nodes to achieve even better spatial reuse. We propose two distributed mechanisms to realize such clustering and show substantial capacity gains over simple random access/ALOHA-like and even RTS/CTS based protocols - on the order of 100-700%. We examine under what regimes such gains can be achieved, and how clustering and contention resolution mechanisms should be optimized to do so. We further extend our MAC design for inducing clustered contention in ad hoc networks to support hop-by-hop relaying on different spatial scales. By allowing nodes to relay beyond the set of nearest neighbors using varying transmission ranges (scales), one can reduce the number of hops between a source and destination so as to meet end-to-end delay requirements. To that end we propose a multi-scale MAC clustering and power control mechanism to support transmissions with different ranges while achieving high spatial reuse. The considerations, analysis and simulations included in this thesis suggest that the principle of inducing spatial clustering in contention has substantial promise towards achieving high spatial reuse, QoS, and energy efficiency in spread spectrum ad hoc networks. |
