Get Complete Project Material File(s) Now! »
Problem addressed in this thesis
Most of the robust tools, like video decoders using JSCD techniques, are not compatible with the SPS due to several non-compliant requirements:
1. The SPS prevents corrupted packets to reach the APL layer, e.g., current WiFi protocol drops all erroneous packets and WiMAX drops all packets with erroneous headers. In the SPSs of both WiMAX and WiFi standards, the retransmission control policy is used to reduce end-to-end packet loss.
2. They require the exchange of soft information between the protocol layers. The SPSs both WiMAX and WiFi do not function with the soft information and hamper the exchange of soft information between the channel decoder at PHY layer and the robust source decoder at APL layer.
3. The conventional use of retransmission mechanism in WiMAX and WiFi, where
erroneous packet is dropped and is retransmitted, is not compatible with these robust decoders.
The solutions of the above-mentioned requirements are normally sooth through the following modifications to the SPS:
1. The protocol stack is usually made permeable to transmission errors by making it capable of forwarding erroneous packets to the APL layer [1]. Thus, the main objective of lower layers is to relay maximum number of packets to the APL layer.
This problem of developing PPS is partially addressed for UDP layer in [2; 3] and for MAC layer in [4; 5; 6], where the ideas of selective error detection are presented.
The adoption of selective error detection, where the packet discard is avoided as long as errors do not affect important bits (e.g., header) of a packet, can produce several positive effects on the network. Nevertheless, the packet drops due to erroneous headers can still become significant, especially at high date rates [7] or in situations where payload-to-header ratio is low. This shortcoming can be addressed partially the receiver based schemes that in addition to ignoring the payload errors can estimate the corrupted header fields [8; 9; 10; 11], thus even packets with erroneous headers can be forwarded to the higher layers.
2. The soft information at PHY layer [12] can reach the APL layer, e.g., by using the transparent layer mechanisms proposed in [13; 14], and thus can be utilized by the robust video decoder. This modification requires changes in the receiver only, and is therefore an applicable solution: it is compliant with the signal which is actually transmitted.
3. A packet received in error needs not be retransmitted unless the robust video decoder at APL layer cannot recover the error. However, these solutions might be difficult to use under several circumstances. The two important complications, one can face while developing a robust video receiver, are studied in this thesis and are mentioned below.
First, often at a given layer of the SPSs of WiMAX and WiFi standards, small packets are aggregated into larger packets or bursts, which are then forwarded to lower protocol layers at the transmitter. This packet aggregation is useful in situations where each transmission unit may have significant overhead (preambles, headers, etc) or where the expected packet size is small compared to the maximum amount of information that can be transmitted.
Thus, it improves the payload-to-header ratio and boosts the throughput of these wireless networks. Similarly, as proposed in [15; 16], the multiplexer can combine multiple packets into a single multiplexed packet and the access point multicasts the multiplexed packet to the wireless end stations to reduce overhead and increase capacity. However, when transmission over a noisy channel is considered, FS, i.e., recovering the aggregated packets from a burst or multiplexed packet, may become difficult. At the receiver, without a robust FS, header estimation techniques mentioned in the solution (1) above remain fruitless, as for them locating the headers and their boundaries is compulsory. Furthermore, FS failures at a lower protocol layer can cause loss of several consecutive packets, which could otherwise be relayed to the APL layer for JSCD. Though,
in case of e.g., WiMAX-link degradation, the user can request the change of modulation and coding scheme to more robust one, it may not be possible in broadcast scenario. Furthermore, some of the state-of-the-art FS techniques [17; 18] are not compliant with the solution (2), as they work on hard bits. Though, several FS techniques, usually deployed at PHY layer, can work on the soft output of the channel [19; 20; 21; 22; 23], they require insertion of the synchronization markers (start codes), which require modifications to the transmitter and an additional overhead, thus cannot be allowed in the protocol stacks of WiMAX/WiFi standards.
Second, for improving the performance, these wireless networks, especially WiFi, tend to include a FEC algorithm [24] to avoid link-layer retransmissions, in situations where the consecutive packets are likely to be infected with bursty error, and in broadcasting applications where retransmission is very difficult if not forbidden. Thus, one needs to provide an additional protection to the packets, so that a lost packet can be recovered at the receiver end. Packet-level FEC decoding to recover a lost packet is often performed on the hard bits, thus necessitating the loss of soft information from the soft-output decoders functioning at PHY layer. Therefore, it hinders the flow of soft information to the higher layers, thus preventing the use of the solution (2). For example, the packet-level FEC protocol RFC 5109 [25] describes method to recover a packet from the uncorrupted hard packets. Furthermore, generally packet-level FEC decoding requires error-free packets to fully retrieve the lost or dropped packets [26], which makes JSCD useless by not allowing erroneous packets to reach the APL layer, thus conflicting with the solution (1).
The induction of contention-free multiple access schemes in IEEE 802.11e [27] and the development of the robust JSCD recovery techniques at various protocol layers have almost eliminated the packet drops in the channel and at the protocol layers, respectively. Therefore, instead of the solution (3), the development of a decoding technique, capable of performing FEC decoding on the soft-valued partially corrupted packets, is the need of the time. Thus, for the broadcast services the retransmissions can be disabled completely and replaced with the packet-level FEC.
Part I: Robust Frame Synchronization
Keeping in view the above-mentioned requirements, in the first part of this thesis, we propose several JPCD approaches for FS. They exploit all available information: soft information at the output of the channel (or channel decoder) as well as the structure of the protocol layers to estimate the boundaries of the small packets and the content of their headers.
First, a trellis-based technique for FS is proposed, where the packet aggregation is modeled by a Markov process, which allow representing all possible successions of packets in a burst by a trellis inspired from that of [28]. A modified BCJR algorithm [29] is applied on this trellis to obtain the packet boundaries. Second, a low-delay and reducedcomplexity suboptimal version of the trellis-based algorithm is proposed. It uses a Sliding Trellis (ST)-based approach inspired from [30], where a low-latency variant of the BCJR algorithm was presented for the decoding of the CCs. These are hold-and-sync(hronize) techniques, which require the whole (for trellis-based) or part (for ST-based) of the burst to perform FS.
Finally, an on-the-fly technique is proposed, which combines robust header recovery technique inspired from [8] with Bayesian hypothesis testing inspired from [19; 20; 21; 22] to localize packet boundaries via a sample-by-sample search. We use a robust 3S automaton, derived from that of [17], but instead of hard CRC correction, a soft header recovery technique [8] for correcting the damaged headers (exploiting all known intra and interlayer redundancies) is exploited to estimate the length field of the header. Moreover, the Bayesian hypothesis testing, used to search for the correct FS, provides improved performance due to the use of soft channel information combined with a priori information due to the redundancy present at the header of a packet.
The FS techniques presented here do not require any signaling overhead, i.e., no synchronization markers are added and only the available information in the protocol layer is utilized. Furthermore, these techniques are quite general, they are illustrated with the synchronization of WiMAX MAC packets aggregated in bursts, which are transmitted to the PHY layer [32], but they are easily extendable to other protocols where packet aggregation is performed.
Part II: Robust Packet-Level FEC Decoding
To broadcast the multimedia packets over WiMAX/WiFi-link, a packet-level FEC scheme is analyzed to overcome the retransmission delays. Taking into account the requirements mentioned earlier, instead of performing decoding on hard bits, the soft information forwarded by lower layers is used to recover erroneous packets. The packet-level FEC decoder is deployed at the RTP layer, where it is assumed that the RTP packets reaching the FEC decoder are soft-valued and can have errors. The same idea of JPCD as deployed in FS techniques is put at work to develop a packet-level Maximum A Posteriori (MAP) decoder to estimate the erroneous packets, utilizing the RTP header redundancies, the redundancy introduced due to use of FEC, and likelihood from the channel. More notably, it causes no hindrance to the flow of soft information from the lower layers, through the RTP layer, to the higher layers. Moreover, the robust decoder presented here needs no
side information and remains completely compatible with RFC 5109.
Though the use of FEC redundant packets would decrease the system goodput, but given that the retransmission is disabled, one can utilize this spared goodput for the transmission of the FEC redundant packets. Based on the channel condition and service nature (real-time video, data, voice, etc.), transmitter can decide for each application or service flow, either to use retransmission or FEC scheme. Transmitter switches to FEC scheme if the channel condition falls below a certain level or in situations when retransmission is causing goodput loss or unacceptable delay.
The FEC protection and the JPCD-based FEC decoder presented in this thesis are well tuned for RTP layer, but they can be extended to the other layers of the S-PPS. Furthermore, they remain compatible with other broadcast scenarios, e.g., of receiving the Mobile TV over DVB-H (Digital Video Broadcasting – Handheld) [33; 34] and then rebroadcasting it over the WiFi network.
Robust tools presented in this thesis, significantly reduce the amount of packets that need to be dropped and enable flow of soft-packets (s-packets), which may contain errors, to the upper layers and enhance the performance of the robust tools functioning at higher layers. They can then be forwarded to the APL layer using the PPS techniques presented in [13; 14; 8], and robustly decoded using JSCD techniques [35; 36].
Table of contents :
Summary
Publications
Contents
List of Figures
List of Tables
1 Introduction
1.1 General Introduction
1.2 Problem addressed in this thesis
1.3 Proposals
1.3.1 Part I: Robust Frame Synchronization
1.3.2 Part II: Robust Packet-Level FEC Decoding
1.4 Thesis Outline
1.4.1 Part I: Robust Frame Synchronization
1.4.2 Part II: Robust Packet-Level FEC Decoding
2 Scenario Under Investigation
2.1 Introduction
2.2 Introduction to WiMAX
2.2.1 WiMAX MAC Layer
2.2.2 WiMAX PHY Layer
2.3 Introduction to WiFi
2.3.1 EDCA
2.3.2 HCCA
2.4 WiMAX-WiFi Inter-networking Scenario
2.4.1 SPS of WiMAX BS
2.4.1.1 Video over WiMAX
2.4.2 SPS of WiBOX
2.4.2.1 802.16 PHY Layer
2.4.2.2 802.16 MAC Layer
2.4.2.3 QoS-Aware Bridge
2.4.2.4 IEEE 802.11 MAC Layer
2.4.3 SPS of WiFi SS
2.4.3.1 IEEE 802.11 PHY Layer
2.4.3.2 IEEE 802.11 MAC Layer
2.4.3.3 UDP Layer
2.4.3.4 APL Layer
2.5 Scenario Considered
2.5.1 S-PPS of WiBOX
2.5.1.1 802.16 Soft-PHY Layer
2.5.1.2 802.16 Soft-MAC Layer
2.5.1.3 RTP-FEC Layer
2.5.1.4 IEEE 802.11 MAC Layer
2.5.2 S-PPS of WiFi SSs
2.5.2.1 IEEE 802.11 Soft-PHY Layer
2.5.2.2 IEEE 802.11 Soft-MAC Layer
2.5.2.3 Soft-IP Layer
2.5.2.4 Soft-UDP Layer
2.5.2.5 RTP-FEC Layer
2.5.2.6 APL Layer
3 Introduction to Frame Synchronization
3.1 Introduction
3.2 Packet Structure
3.3 Fixed-length State-of-the-art FS techniques
3.3.1 Cell Delineation
3.3.1.1 Modified CD
3.4 Variable-length State-of-the-art FS techniques
3.4.1 NP FS Method
3.4.2 Ueda’s Method
3.4.2.1 Modified Ueda’s (MU) method:
3.5 Simulation results
3.6 Conclusions
4 Trellis-based FS
4.1 Introduction
4.2 Header structure
4.3 Trellis Representation of a burst
4.4 MAP estimation for FS
4.4.1 Estimators for the number of packets and their boundaries
4.5 Trellis-based FS Algorithm
4.5.1 Evaluation of n and n
4.5.2 Evaluation of
4.5.2.1 First case, ` < L
4.5.2.2 Second case, ` = L
4.5.3 Complexity evaluation
4.5.4 Limitations
4.6 Simulation results
4.7 Conclusions
5 Sliding Trellis-based FS
5.1 Introduction
5.2 Sliding Trellis-based FS Algorithm
5.2.1 Sliding Trellis
5.2.2 Evaluation of
5.2.3 Initialization of in the sliding trellises
5.2.4 Complexity Gain
5.3 Simulation results
5.4 Conclusions
6 Robust Three-State FS Automaton
6.1 Introduction
6.2 Robust 3S FS Automaton
6.2.1 SYNC State: Header Recovery
6.2.2 HUNT State: Bayesian hypothesis test
6.2.2.1 A priori probabilities
6.3 Simulation results
6.4 Conclusions
7 Introduction to Packet-Level Forward Error Correction
7.1 Introduction
7.2 System Model
7.3 A Simple RTP-Level FEC Scheme
7.3.1 FECP Construction
7.3.1.1 RTP header fields of FECP.
7.3.1.2 FEC header fields of FECP.
7.3.1.3 FEC-level header fields of FECP.
7.3.1.4 Payload of FECP.
7.4 Theoretical Analysis
7.4.1 Retransmission Scenario
7.4.2 Packet-level FEC Scheme
7.5 Simulation results
7.6 Conclusions
8 Robust MAP Decoding for RTP-Level FEC
8.1 Introduction
8.2 MAP Estimation for Packet-level FEC
8.3 Packet Structure
8.3.1 MedP Structure
8.3.2 FECP Structure
8.4 MAP Decoding for RTP-Level FEC
8.4.1 MAP Decoding for RTP Header
8.4.2 MAP Decoding for RTP Payload
8.5 Simulation results
8.6 Limitations
8.7 Conclusions
9 Conclusion
9.1 Contributions
9.2 Future Work and Applications
9.2.1 FS Applications
9.2.1.1 SDL Framing
9.2.1.2 IEEE 802.11n Packet Aggregation
9.2.2 Packet-Level FEC Decoder Applications
A Appendices of different Chapters
A.1 WiMAX PHY FRAME
A.1.1 OFDM PHY DL Sub-frame
A.1.2 OFDM PHY UL Sub-frame
A.2 WIMAX MAC PDU
A.2.1 Generic MAC Header (GMH)
A.2.2 The Bandwidth Request Header (BRH)
A.3 UDP Header
A.4 UDP-Lite Header
A.5 RTP Header
Bibliography