A connection-level call admission control using genetic algorithm for multi-class multimedia services in wireless networks

Call admission control in a wireless cell in a personal communication system (PCS) can be modeled as an M/M/C/C queuing system with m classes of users. Semi-Markov Decision Process (SMDP) can be used to optimize channel utilization with upper bounds on handoff blocking probabilities as Quality of Service constraints. However, this method is too time-consuming and therefore it fails when state space and action space are large. In this paper, we apply a genetic algorithm approach to address the situation when the SMDP approach fails. We code call admission control decisions as binary strings, where a value of “1” in the position i (i=1,...m) of a decision string stands for the decision of accepting a call in class-i; a value of “0” in the position i of the decision string stands for the decision of rejecting a call in class-i. The coded binary strings are feed into the genetic algorithm, and the resulting binary strings are founded to be near optimal call admission control decisions. Simulation results from the genetic algorithm are compared with the optimal solutions obtained from linear programming for the SMDP approach. The results reveal that the genetic algorithm approximates the optimal approach very well with less complexity.