Achieving reliability for wireless multicast transmission using network coding
Date of Issue2014
School of Computer Engineering
Symbol erasure is one the fundamental and inevitable characteristic in data transmission network. This problem exuberates for data transmission on wireless channel due to the shared medium of transmission resulting in packet collisions. Achieving network reliability on wireless channel is even more difficult when data is transmitted to multiple receivers i.e. for wireless multicast transmission. An efficient solution to achieve reliability for wireless multicast transmission is clearly extendable for relatively simpler data transmission models, such as wireline transmission network, wireless unicast transmission, and even on data storage systems to deal with symbol erasures. In this thesis we attempt to address the reliability of wireless multicast transmission at two layers. In the first layer we address the characteristics of the code design to retransmit erased packets. The metrics which we optimize for our code design include throughput performance, decoding delay and encoding and decoding complexities. In the second layer we propose physical layer network coding (PNC) based collision codes to scalably collect packet feedback information from multiple receivers. We further show that PNC based transmission scheme can be harvested for interfering wireless multicast networks to improve the aggregate retransmission throughput performance of the interfering networks. The solutions proposed at each of these two layers complement each other to improve the overall reliability of wireless multicast transmission. An efficient coding decision is dependent on the information about the packet reception status at the receivers. We first develop a scalable scheme to collect packet acknowledgement frames by designing a collision coding scheme whereby all the receivers can simultaneously transmit their acknowledgement frames, resulting in transmission collision. Based on the collided signal, and making use of the collision codes, the transmitter can decode information about the set of receivers which have transmitted their acknowledgement frames. Given the packet feedback information at the transmitter we then propose a coding algorithm, which we call BENEFIT, to design GF(2) linear codes to recover the lost packets. We show that our proposed code has the best throughput and decoding delay performance of any GF(2) linear codes. We then show that by taking advantages of opportunistic listening due to the shared medium of wireless transmission, overlapping transmission range of the interfering transmitters and physical layer network coding scheme, the results of collision codes and BENEFIT coding algorithm, can be extended to interfering multicast networks to improve the aggregate retransmission throughput performance. In the last part of our thesis we design efficient erasure codes where the coding decision is made independently of the feedback information from the receivers, thus completely eliminating the overhead of feedback transmissions from the receivers. The main highlight of these codes, which we call triangular codes, is that triangular codes are the first class of erasure codes which can achieve near-optimal transmission rates, with linear encoding and decoding complexities for finite length input symbol length. We further show that triangular codes outperform all erasure codes with quadratic or subquadratic decoding complexities, including all versions of Luby-Transform codes, Raptor codes, and the standardized versions of Raptor codes, the Raptor 10 and Raptor Q codes.
DRNTU::Engineering::Computer science and engineering