Optimizing Throughput with Carrier Sensing Adaptation for IEEE 802.11 Mesh Networks Based on Loss Differentiation Hui Ma, Soo Young Shin, Sumit Roy Dept. of Electrical Engineering University of Washington, Box 352500 Seattle, WA 98195-2500 {mahui,wdragon,roy}@ee.washington.edu Abstract— In high density (HD) mesh networks, packet losses can occur due to co-channel interference (asynchronous inter- ference) or collisions (synchronous interference). In this paper, we first propose a novel method of estimating the probability of collision and interference statistically. Further, we integrate this differentiation method with physical carrier sensing adaptation in a novel centralized algorithm to improve the aggregate through- put in HD mesh network. Extensive simulations results show that the on-line algorithm approaches the optimal throughput predicted by analytical models. Index Terms— Physical Carrier Sense, Mesh Networks, Ad- Hoc, Contention Window Size, Adaptation I. I NTRODUCTION Wireless networks at the ‘edge’ are proliferating both in numbers and scale. In the next generation networks, ad hoc mesh (or multihop) networks are likely served as an interme- diary that provides broadband connectivity to the backbone Internet for mobile client devices in various environments such as campus, office and home. The proliferation of consumer electronic and wireless-enabled mobile computing devices will continually increase the node density that must be supported by such mesh networks. In such dense environments, exploit- ing the limited system bandwidth available via spatial reuse becomes a key to improving the aggregate network throughput. Physical carrier sensing (PCS) with tunable thresholds (PCS threshold) is proven to be an efficient method for managing mutual interference from simultaneous co-channel transmis- sions in a mesh network [1], [2], [3]. Here, each node samples the energy level in the medium and initiates channel access only if the received signal strength is below the PCS threshold. PCS with variable threshold has been shown to yield better aggregated throughput comparing to the static PCS threshold in IEEE 802.11 [11]. Zhu et al. [1] derived the optimal PCS threshold that maximizes the aggregate one-hop throughput for a regular topology given a minimum required SNR; an adap- tive PCS threshold algorithm was suggested based on periodic measurement of Packet Error Rate (PER) and evaluated on a real test-bed in [2]. A novel analytical model was introduced This work was supported in part by Intel Corpn. and T-Mobile USA. in [3] for determining the optimal carrier sensing range by minimizing the sum of the hidden terminal event and exposed terminal event. It was shown that an optimal PCS threshold achieves a trade-off between the amount of spatial reuse and the PER due to hidden terminals, thereby improving the overall network throughput. Hence, real-time measurement of the effect of hidden terminals is the key towards design of effective adaptation algorithms in high density (HD) mesh networks. In a HD mesh network, the probability of collisions (apart from asynchronous interference) is significant, because of potentially simultaneous transmissions starting in the same time slot. Any adaptation scheme that does not consider these collisions can lead to lower-than-optimal aggregate throughput or even diverge in extreme conditions. Therefore, in the design of adaptation algorithm for HD mesh, determining the cause of the packet losses (e.g. differentiation of measured PER into those result- ing from hidden terminals and collisions, respectively) is one of the primary challenges. The above is particularly difficult because the typical re- sponse to a packet transmission is coarse (binary): in ACK based systems, the transmitter only knows success/failure and not the cause of losses. There has been several at- tempts to distinguish the cause of packet losses in wireless networks. For example, [4] relies on request-to-send/clear- to-send (RTS/CTS) exchange in 802.11 for differentiation. However, RTS/CTS suffers from inefficiency and fundamental limitations in spatial reuse [1] [8]. In [5], all stations broadcast their transmission time for the failed transmissions, which con- tributes to additional communication and processing overhead. Exploiting the capture effect was proposed in [6] for detecting collisions, which requires re-designing the receive chain in the IEEE 802.11 hardware. In [7], although a novel loss differentiation MAC was proposed, it required a new MAC frame, thereby compromising compatibility with existing IEEE 802.11 standard. Therefore, there continues to exist a need for low-overhead, robust yet accurate loss differentiation method for IEEE 802.11 HD mesh networks. In this paper, we contribute a 1-4244-0353-7/07/$25.00 ©2007 IEEE This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the ICC 2007 proceedings.