On the Error Exponent and the Use of LDPC Codes for Cooperative Sensor Networks With Misinformed Nodes

The problem of retrieving information by a mobile access point from a sensor network where sensors cooperatively transmit messages using a common codebook is considered. It is assumed that there is a probability that a sensor is misinformed with a wrong message, which complicates the information retrieval process. The access point uses the capacity achieving stay-k scheduler that schedules a sensor to transmit for k consecutive code-letters before switching to a new sensor. The random coding exponent is derived as a function of k, and it is shown that there is an optimal k that gives the largest error exponent. The application of low-definition parity-check (LDPC) codes is considered next. It is shown in simulations that the optimal k of the stay-k scheduler for LDPC codes can be inferred from that for the random coding exponent.