Research issues recently proposed of 60 GHz communication include physical implementation and system architecture design, beamforming technical design, effective performance analysis and improvement, relay selection, and resource allocation. Our thesis focuses on the resource management and relay selection between multi-piconet communications.
In IEEE 802.15.3 standard, the PNC uses conventional scheduling scheme that allocating a unique CTA to one data flow and the interference-free transmission is guaranteed. The scheduling scheme does not benefit from the usage of beamforming technology and spatial reuse so the system throughput in 802.15.3 WPAN has room for improvement. An enhanced IEEE 802.15.3 traditional scheduling scheme to support simultaneously operating piconets (SOPs) is proposed in [8]. The throughput is limited because the same superframe is shared by the dependent piconets as we mentioned in chapter 1; therefore, they defined the CTAs adaptively as normal CTAs and public CTAs that mitigate the inter-piconet interference and enhance the efficiency. Against to traditional superframe sharing configuration, each dependent piconet maintains its own superframe. To avoid beacon collisions in the overlapped area, the timing of all existing superframes are informed by operating piconet coordination and beacon alignment. As for interference occurred in CTAP, the PNCs schedule their intra-piconet flows into the normal CTAs, and these flows can transmit simultaneously; the PNCs allocate the public CTAs interlaced to transmit the inter-piconet flows. However, blocking the time interval in one piconet while the inter-piconet flows transmit in the other piconet is not efficient, so we do some modification that two flows can be scheduled in a CTA if no interference happened regardless of the flow types.
8
A randomized exclusive region (REX) based scheduling was introduced in [9, 10]
to explore the spatial multiplexing gain in mmWave WPANs. The paper derived the exclusive region (ER) based on the use of omni- or directional antennas to allow concurrent flows transmission that are favorable in terms of per flow throughput and network throughput. Every receiver of a flow can draw its ER by computing a mainlobe and sidelobe of gain. If two flows are mutually outside each other’s exclusive regions, the flows can be scheduled to the same time duration to transmit. In REX, one flow is selected to be scheduled in a time slot; then, all remaining flows are verified according to the ER condition to discover spatial channel reuse-capable flows.
The paper proposed an effective scheduling idea, but the PNC has to measure the channel gain to manage those flows. Though concurrent transmission is considered, inter-piconet communications and relay selection are not addressed.
Effective throughput of mmWave WPAN deploying relay was studied in [11].
Taking the impact from concurrent transmission into consideration, they proposed a deflection routing algorithm to maximize the effective throughput. By collecting the co-channel interference (CCI) periodically, the PNC maintains a table consisting of all the potential CCI to and from each and every link, and the PNC determines the suitable relay node and timing for requesting flows. In [12], the authors further formulated the scheduling problem and transfer this problem as a max-weight matching problem of a bipartite which can be solved by Kuhn-Munkres algorithm.
However, devices send periodical probing signals so that the overhead gained inevitably. The contributions of these papers only work on the intra-piconet flows, and they do not feature the characteristics of directional antenna. [13] classified communications into two categories: direct path and relay path, and propose a deflection routing scheme. Both share the channel resource when signal-to-noise-plus-interference ration (SINR) of a receiver is acceptable. This
9
method improves the effective throughput by sharing time slots for direct path with relay path. The authors also propose Virtual Time-Slot Allocation (VTSA) to allow multiple flows concurrently transmitting in per slot basis. By maintaining the CCI table, the PNC assign the flows with less CCI value to share the same CTA. The CTA sharing the same channel resource with primary CTA is named a virtual CTA. The CTA is reusable and the throughput is improved. PNCs collect the information of co-channel interference to do relay selection, while the cost of probing channel statuses is high.
In [14], the authors proposed a joint link scheduling, channel assignment, and routing scheme for the 60 GHz multi-channel wireless mesh networks with directional antennas. They first designed a linear programming framework to model the network throughput of multi-channel multi-radio wireless mesh network and then study the routing optimization problem. It also proposed a heuristic greedy algorithm to obtain a feasible solution to approximate the optimal network throughput. Te greedy algorithm went to estimate the flow rate of each link on a channel in a CTA and schedule the link flow every CTA. Once collecting every possible link and channel information in the mesh network, the PNC will assign first flow to the channel that can provide maximum flow rate and assign next flow with second highest rate that would not interfere to those scheduled flows in the same time slot. This paper considered multiple channel assignment which is not the issue in our research.
Since the duration of superframe is not fixed but limited to 20 ms, the time division also plays a role in improving system throughput. In [15], the paper studied the throughput of the 802.15.3c WPAN system that is based on the hybrid multiple access of CSMA/CA (CAP) and TDMA (CTAP). From the analysis, the authors found that the throughput is significantly affected by the access time of CAP even if data transmission is mainly performed in CTAP. A tradeoff occured while a large access
10
time of CAP could reduce data transmission collisions; however, it would lead to degradation in the throughput since the data transmission time in CTAP was reduced relatively. Otherwise, lower the access time of CAP would not benefit the system throughput. The increasing data transmission collisions along with the shorter CAP would also lead to a reduction in the throughput due to an increase in the unused data transmission time in CTAP because of the failure of data transmission channel release in CAP. The paper provided the approximated optimum access time of CAP with different contending devices for maximizing the throughput. Moreover, it introduced throughput improvement through a reduction in the failures of data transmission channel release. In our simulation part, we set our parameters refer to the analysis result of this paper to gain our throughput.
11