DECENTRALIZED SUBCARRIER AND POWER ALLOCATION SCHEMES

The distributed infrastructure of WiMAX mesh and relay networks, and the need for reducing communication overhead between the BS and network nodes are the motivations behind proposing decentralized resource allocation schemes. Potentially, either no central controller exists or it does not influence the allocation decision in a decentralized resource allocation. These facts make the decentralized schemes more scalable.

Resource allocation in decentralized networks is essentially different from centralized one. In centralized schemes, the BS collects CSI from all users, allocates the subcarriers or power to the users, and informs the users of allocated resources. On the other hand, users may not need to know the CSI of the other parties in decentralized schemes. Besides, the parties may not be aware of the decision of each other, so a collision is probable. Accordingly, in each proposed distributed resource allocation scheme for OFDM-based networks, the following questions should be answered:

• How does a user achieve the required CSI? The hidden terminal and exposed node problems are common problems in self-organized and decentralized networks, because it is assumed that no central controller exists to assist in signaling. Besides, the signaling overhead should be reasonable for a practical implementation.
  
  
• How should the node coordinate or compete with the other nodes to attain the resources? A node does not know the requirements of the other nodes, at each instant, so competence or coordination is a “must” for a node which wants to start capturing the resources.

An interference aware subchannel allocation scheme that overcomes the drawbacks of decentralized schemes. 

As the scheme uses updated CSI at the beginning of each MAC frame, the channel does not need to be assumed time-invariant over multiple MAC frames. The scheme is appropriate for OFDM-TDD networks. The MAC frame is divided into mini-slots; at the first mini-slot of each MAC frame the ongoing receiver nodes broadcast a busy signal to inform the respected transmitters of the quality of the allocated subchannel. The inactive nodes does not send any busy signal. The ongoing transmitter nodes listen to the busy signal and adapt their subcarrier allocation to their specific receiver nodes according to the information on the busy signal. Also, a node that wants to start transmission listens to the busy signal and chooses the subcarriers that are not interfered by ongoing transmissions and their interference. In other words, it selects those subcarriers with a received busy signal power less than a threshold. The advantages of the scheme are as follows:

• Signaling overhead is low compared to other radio resource allocation schemes
• The co-channel interference is reduced significantly since the scheme is interference aware
• Full frequency reuse is possible

A decentralized power allocation problem for a cooperative transmission is formulated in Ref. [43]. The objective is to minimize the total transmission power of all users subject to providing minimum rate requirement of each user. The network model uses a time division multiple access with the OFDM multiplexing (TDMA-OFDM), so only one user accesses the total OFDM subcarriers in each time slot. The user may allocate some of the subcarriers to transmit its traffic and the rest of subcarriers to relay the other users’ traffic to minimize total transmission power. In other words, the resource allocation problem determines if a user should cooperate with other users, and in case of cooperation it determines the subcarriers and power allocation. A cooperative user may be allocated more power than a noncooperative user, because its location and channel gain allows it to cooperate with other users. However, the total network power is minimized by cooperation among users. A decoupled subcarrier and power allocation scheme is proposed to solve the problem.

Assuming an access point in the network proposes a distributed decision-making scheme for the resource allocation. Each user measures its CSI upon receiving a beacon signal from the access point. The subcarriers are divided into several groups (equal to the number of users), and the approximate channel gain of each group for each user is estimated. Then, the users contend with each other to achieve the group with the best channel quality of their own. A backoff mechanism is proposed to avoid collision. Each user start contending for the best group of subcarriers after a backoff time which is proportional to the best group gain for that user. After contention, the access point informs the users of the winners which can transmit in the next transmission interval. A frame is divided into three subframes, contention, acknowledge, and transmission subframes. This scheme is actually a distributed resource allocation scheme for a centralized network that aims to reduce the signaling overhead and processing task of the access point.

A collaborative subcarrier allocation using the swarm intelligent. The scheme relies on the central controller to achieve the updated information of available subcarriers and the highest and lowest demands for each subcarrier. In other words, the nodes negotiate with the central controller iteratively in a negotiation phase. In each iteration, the nodes inform the central controller of their demand for a specific subcarrier. Then, the central controller broadcast a message indicating the highest and lowest demands for each subcarrier. Based on the feedback messages that occur several times in each negotiation phase, the nodes intelligently decide upon subcarriers.

A comparison between the capacity performance of OFDM-TDMA and OFDM-FDMA in a two-hop distributed network. The time frame is divided by two in OFDM-TDMA. In each half of the time frame, all subcarriers are devoted for transmission over one hop. A power allocation algorithm is proposed to maximize the end-to-end capacity subject to the overall transmit power constraint of the two hops. For OFDM-FDMA, the subcarriers are assigned to the two hops without overlapping and a joint subcarrier and power allocation algorithm is proposed. The simulation performance results of the algorithms show that OFDM-FDMA achieves a higher end-to-end capacity than OFDM-TDMA.

MAC SUBLAYER | OFDM-BASED WiMAX

Resource allocation is one of the major tasks of MAC because the medium access mechanisms of MAC directly affect the spectrum and power utilization. Accordingly, the MAC sublayer specifications in the IEEE 802.16 standard, relevant to our discussion in the next section, are introduced in the following.

Despite the advantages of OFDM in mitigating the channels’s impairments as mentioned before, underutilization of transmitter power and network subcarriers is its disadvantage. When an OFDM transmitter accesses the channel in a time division manner, e.g., time division multiple access (TDMA), the transmitter is forced to transmit on all available subcarriers N_sc, although it may require a less number of subcarriers to satisfy its transmission rate requirement. Consequently, the transmitter power consumption increases as the number of subcarriers increases. This disadvantage motivates the development of a PHY technology where transmitters are multiplexed in time and frequency, i.e., OFDMA. In such a technology, the users are exclusively assigned a subset of the network available subcarriers in each time slot. The number of both time slots and subcarriers can be dynamically assigned to each user; this is referred to as dynamic subcarrier assignment which introduces multiuser diversity. The multiuser diversity gain arises from the fact that the utilization of given resources varies from one user to another. A subcarrier may be in deep fading for one user  while it is not for another user (e.g., the same subcarrier for user O). Allocating this particular subcarrier to the user with higher channel gain permits higher transmission rate. To achieve multiuser diversity gain, a scheduler at the MAC sublayer is required to schedule users in appropriate frequency and time slots.

Point-to-multipoint (PMP) as well as mesh topologies are supported in the IEEE 802.16 standard. PMP mode operations are centrally controlled by the BS, but mesh mode can be either centralized or decentralized, i.e., distributed. In a centralized mesh, a mesh BS (a node that is directly connected to the backbone) coordinates communications among the nodes. Decentralized mesh is similar to multihop adhoc networks in the sense that the nodes should coordinate among themselves to avoid collision or reduce the transmission interference.

In PMP mode, the uplink (UL) channel, transmissions from users to BS, is shared by all users, i.e., UL is a multiple access channel. On the other hand, downlink (DL) channel, transmissions from BS to subscriber stations (SSs), is a broadcast channel. The duplexing methods of UL and DL include time division duplexing (TDD), frequency division duplexing (FDD), and half-duplex FDD (HFDD). Unlike PMP mode, there is no clear UL and DL channel defined for mesh mode.

An outstanding feature of WiMAX is heterogeneous traffic support over wireless channels. IEEE 802.16 provides service for four traffic types known as service flows. The mechanism of bandwidth assignment to each SS depends on the QoS requirements of its service flows. The service flows and their corresponding bandwidth request mechanisms are as follows:

• Unsolicited grant service (UGS): This service supports constant bit rate traffic. Bandwidth is granted to this service periodically or in case of traffic presence by the BS.
  
  
• Real-time polling service (rtPS): This service has been provided for real-time service flows with variable-size data packets issued periodically. rtPS flows can send their bandwidth request to the BS after being polled.
  
  
• Nonreal-time polling service (nrtPS): This service is for nonreal-time traffic with variable-size data packets. nrtPS can gain access to the channel using monocast or multicast polling mechanisms. Upon receiving a multicast polling, the nrtPS service can take part in a contention in the bandwidth contention range.
  
  
• Best effort (BE) service: This service provides the minimum required QoS for nonreal-time traffic. The channel access mechanism of this service is based on contention.

WIMAX HARQ

HARQ is supported in WiMAX that uses OFDMA physical layer. As mentioned before, WiMAX uses a basic stop-and-wait HARQ protocol. Using multiple HARQ channels can compensate the propagation delay of the stop-and-wait scheme, that is, one channel transmits data while others are waiting for feedbacks. Therefore, using a small number of HARQ channels (e.g., using 6 channels), multichannel stop-and-wait HARQ is a simple, efficient protocol that simultaneously achieves high throughput and low memory requirement.

In this scheme, each HARQ channel is independent of each other; that is, a data burst can only be retransmitted by the HARQ channel that initially sent the data burst. Each HARQ channel is distinguished by a HARQ channel identifier (ACID – ARQ channel identifier). Data reordering in an SS is done by referring to Protocol Data Unit (PDU) sequence numbers that are enabled when the HARQ operation is used. The HARQ Downlink Map Information Element (DL MAP IE) contains the information about the DL HARQ Chase sub-burst IE. It specifies the location of HARQ sub bursts, the ACID, the HARQ Identifier Sequence Number (AI_SN), etc. By referring to MAP IE, an SS can correctly retrieve a given data burst. HARQ feedbacks are sent by the SS after a fixed delay (this is called synchronous feedbacks). To specify the start of a new transmission on each HARQ channel, one-bit HARQ AI_SN is toggled on each successful transmission.

When CC is used in WiMAX HARQ, the total buffer capability needed in a specific SS is determined by the maximum data size and may be transmitted by a single channel at a time, and the total number of HARQ channels provided for the SS (i.e., the product of multiplying these two values). In the IEEE 802.16 Standard, this buffer capacity is used by the BS as a rate control mechanism when sending data to an SS.

1. A Simple Example for WiMAX HARQ

In this section, the multiple-channel stop-and-wait HARQ used by WiMAX is illustrated by a simple example. As mentioned above, these HARQ channels are independent of each other; retransmissions of a data burst can only be done by its initial sending HARQ channel. Thus, a negative acknowledged (NAKed) data burst can only be resent via the initial sending channel until it is successfully received.

Figure 1 shows the channel activities versus time frames of a HARQ scheme, from a BS to a particular SS, assuming that four channels are available for the SS. The first row shows the consecutive time frames. The next four rows show the data bursts sent and the ACK or NAK received from/by each channel at the corresponding time frame. The last row shows the data bursts storing in the SS buffer at the corresponding time frame (waiting to be forwarded to the upper layer). It is assumed that data bursts have either been received correctly or erroneously. Erroneously received data bursts are placed in the receiver’s buffer waiting for the correct copies. On the other hand, correctly received data bursts are placed in the receiver’s buffer waiting to be forwarded further (either to the next hop or to the upper layer) in the right sequence, that is, the receiver is still waiting for a correct copy of some earlier data bursts.

 

Figure 1: A simple example of the WiMAX HARQ scheme.

Figure 1 may be explained below. Data bursts 1 through 4 are sent via channels 1 through 4, at time frames 1 through 4, respectively. A corresponding ACK or NAK will be received by the BS by the end of the 4th time frame after the BS sent out the data burst. So, at the end of 4th time frame, a NAK for data burst 1 is received, and thus the second copy of data burst 1 is sent at time frame 5, via channel 1 (the same channel it was initially sent). Similarly the second copy of data burst 2 is sent at time frame 6 via channel 2. Since an ACK for data burst 3 is received at time frame 6, a new data burst (5) is sent via channel 3 at time frame 7.

The data bursts stored at the receiver’s buffer may be explained as follows. At time frame 4, data bursts 1, 2, and 3 are in the receiver’s buffer. Data bursts 1 and 2 have been erroneously received (and are waiting for correct copies) while data burst 3, even though has been correctly received, is also waiting so that it may be forwarded further in the correct order. Similarly, at time frame 5, data bursts 1 through 4 are in the buffer. At time frame 6, only data bursts 2 through 4 are left in the buffer since data burst 1 has been correctly forwarded. At time 7, after data burst 2 is correctly forwarded, data bursts 3 and 4 are also forwarded, so nothing is left in the buffer. The rest of the time frames may be explained similarly.