SPLITTING ALGORITHMS IN CHANNELS WITH MARKOVIAN CAPTURE

In this paper we study the performance of tree-like splitting collisio n resolution algorithms in channels with markovian capture. In particu lar, we assume that in each slot the channel can be in one of two stat es -b (for ''bad'') and g (for ''good''). When the channel is in state b, a capture can never occur. When the channel is in state g and n no des (n greater-than-or-equal-to 2) are transmitting, a capture occurs with probability pi(n). The sequence of channel states is assumed to b e a homogeneous Markov chain. We derive the throughput of a splitting tree-like multiple access algorithm for this channel. We also provide simulation results for the average delay.