Teknillinen tiedekunta, 2011
Prof. Mohammed Elmusrati
Master's Degree Programme in Telecommunication Engineering
To improve reliability and save precious energy resources of wireless sensor networks, a cross-layer optimization scheme that optimizes retransmission requests was investigated. We have worked to achieve a better error performance which reduces the amount of Automatic Repeat Requests (ARQs) inside wireless sensor networks by taking into account local recovery inside the decoder for a packet which was determined to be in error after forward error correction (FEC).
Due to channel noise and fading, error in a received vector can cause selection of wrong codeword other than the transmitted one. This is normally what causes a decoder error. Conventionally, after the best effort of FEC, a packet which fails upper layer error detection will be automatically dropped and retransmission is requested. Traditional decoders have ignored the probability of other valid codewords to be the right one and retransmission is done for the same packet which would otherwise be recovered locally. We propose local recovery of wrong packets by considering the probability of other valid codewords to be the correct one. We determined the likelihood of each codeword to be the right one based on how close they are from the received vector. This way the synergy effect of the first and second best likelihood codewords on the decoder error performance has been tested. The test was first done by simple block code, which has one parity check bit for error detection, and later with more practical decoder, the viterbi decoder. A convolutional code of rate 1/3 and constraint length K of 3 was considered for the latter.