Stochastic estimator-based wireless traffic control schemes

Abstract

Recent works on Available Bit Rate (ABR) traffic control have generated efficient control schemes for ABR traffic on Asynchronous Transfer Mode (ATM) network. This study examines the improved performance envisaged if these control schemes adjust dynamically to the varying ABR bandwidth capacity in a stochastic manner instead of conventional deterministic approach. The performance difference between setting explicit rate deterministically for transmitting ABR sources and doing the same stochastically using a learning automaton is of particular interest. The learning automaton used is the Stochastic Estimator Learning Automaton (SELA). The performance difference is measured by comparing the congestion levels of the SELA-based control scheme with the reference deterministic control mechanism. Simulation results show that the stochastic estimator gives a better performance. The higher average congestion level experienced by the conventional deterministic approach is mainly due to the propagation time delay in the closed-loop feedback control schemes.