Dynamic control of time slot assignment in multiaccess reservation protocols

The authors consider the dynamic control of time slot assignment in reservation protocols for multiaccess channels employing frame structures similar to those investigated by Roberts (1973) and others. Such protocols are particularly suitable for channels with long propagation delays, such as satellite channels. Fixed length frames are dynamically divided on a frame by frame basis into reservation subframes with minislots for reservation requests transmitted by slotted ALOHA random access, and data subframes with reserved time slots for data packet transmissions. Formulating the problem as a Markov decision process, the authors derive the optimal time slot assignment policy, which makes use of complete state information of the finite user population, and obtain the delay-throughput performance. Results show that the optimal policy provides substantial performance improvements over fixed assignment policies. Furthermore, a heuristic dynamic assignment policy called the MF policy is proposed that is much easier to derive and implement, particularly for large networks, while maintaining performance very close to the optimal. Both of these dynamic policies depend on system state information, which the authors estimate using a generalised pseudo-Bayesian estimator. The use of this estimator results in only a small reduction in performance