Classification-Based VQ Model for Simulation of Binary Error Process on the Wireless Channel

Realistic binary error process models are particularly important for proper design of throughput increasing techniques, such as Forward Error Correction (FEC), Automatic Repeat Request (ARQ), hybrid ARQ or cross-layer optimization. Currently employed models are capable of producing realistic output for binary error process observed on the error control channel, however, most of them (including standard Gilbert and generalized Elliot's model) encounter significant limitations when modeling the binary error process of the logical channel. The proposed novel classification-based vector quantization (VQ) model is designed specifically to produce realistic error burst and error gap process of the binary error process regardless of binary channel type, also retaining high precision of the overall cluster error probability characteristic. The proposed model's precision was verified using standard statistical distances against a real wireless sensor network logical channel trace. Furthermore, the Pearson's goodness of fit test was used to verify viability of the proposed model variants. Multiple variants demonstrate both high precision and successfully pass the goodness of fit test.