| Popis: |
Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA) is still the de facto contention-based Medium Access Control (MAC) protocol in many of today’s Wireless Local (WLAN) and Personal Area Network (WPAN) standards such as the IEEE 802.11a/b/g/n/ad, IEEE 802.15.3, IEEE 802.15.4, and ECMA 387. While CSMA/CA is efficient in single cell scenarios, in multi-cell scenarios it suffers a severe MAC-level co-channel interference problem which affects its spatial reusability. Known as the exposed node problem, it prevents nodes in different cells within carrier-sensing range from sharing a channel even though the cells are interference-limited. To mitigate this problem, many schemes have been proposed, of which schemes combining carrier sensing threshold (CST) and transmit power control (TPC) stand out in terms of practicality. Unfortunately, such schemes require features that may not be available in simpler transceivers. Also, for CST to work, transceivers must use a carriersensing mode that can lead to poor detection rates especially for wideband signals. In this thesis, we propose an alternative scheme to overcome this issue: the Payload-Dropping CSMA/CA (PD-CSMA/CA). This special variant of the CSMA/CA protocol incorporates a MAC/PHY cross-layer mechanism which aborts the reception of the payload portions of frames from co-channel cells in interference-limited multi-cell deployments, based on a cell identifier embedded in the frame’s header. For transceivers which do not support CST, PD-CSMA/CA adequately mitigates the exposed node problem, thereby allowing nodes to enjoy nearly the maximum throughput as provided by CSMA/CA in the single cell scenario. To evaluate this new scheme, we incorporated it into the IEEE 802.11’s CSMA/CA protocol and tested its performance in three different indoor spatial re-use scenarios using the ns-2 simulator. From the simulation results, it can be seen that in deployments where either cell spacings or partitions are used to limit the co-channel interference, better throughputs are achieved when PD-CSMA/CA is used instead of CSMA/CA. The results also show that, under exactly the same deployment scenario and propagation environment, and using exactly the same TPC and CST settings, the throughput for PDCSMA is more than the throughput for CSMA/CA, and the increase in the throughput for PD-CSMA/CA is larger than the increase of the throughput for CSMA/CA as the fade margin (employed to combat lognormal shadowing) for both protocols is increased in equal amount. We also developed analytical formulations for the throughput of PD-CSMA/CA in two co-channel interference-limited cells using the Markov chain modeling approach. For comparison, the throughput model for CSMA/CA in the same setup was also developed. Compared to simulations results, these models are very accurate. Although limited to the fixed frame size and fixed contention window case, these models shed light on the throughput trends of PD-CSMA/CA with respect to the contention window length, number of nodes and header to frame length ratio of transmitted frames, and demonstrates its throughput gains over normal CSMA/CA theoretically. As a by-product of the effort to develop the analytical model for PD-CSMA/CA, we also developed formulations for the idle period distribution of single-cell CSMA/CA. This model passed the Pearson’s Chi-squared test for a wide range of contention window sizes and numbers of nodes. Although limited to fixed contention windows, we feel it is an important first step towards developing a more general expression for CSMA/CA with exponentially increasing windows. DOCTOR OF PHILOSOPHY (SCE) |
