Acceleration of performance evaluations for adaptive coding protocols in frequency-hop communications

Effective adaptation of error-control coding in a packet radio network requires certain statistics to be developed in the demodulator and decoder of the receiving radio during the reception of each packet. The statistics from one packet are employed to select the code for the next packet. The design and evaluation of cross-layer protocols that use such statistics typically require numerous simulations of the network, and each simulation of the network has numerous embedded simulations of the demodulators and decoders that are in the radios. The embedded simulations of the decoder are especially computationally intensive if the radios employ iterative decoding. We present a method for the direct generation of receiver statistics that avoids the time-consuming embedded simulations of the demodulator and decoder, thereby accelerating the process of designing, developing, and evaluating adaptive cross-layer protocols. The required statistics are modeled as random variables that can be generated quickly, yet their distributions approximate the distributions of the actual receiver statistics. The method is illustrated for an adaptive coding protocol in a network of frequency-hop (FH) radios that employ turbo product codes with soft-decision iterative decoding. Results are given on the quality of the approximations and the reductions in simulation time that they provide. Our method for the direct generation of receiver statistics provides even greater benefits in the design and evaluation of higher-layer protocols that employ receiver statistics.