A simulated annealing applied for optimizing a voice-multihop radio network (Q2570727)

In the article the optimization of admission control in a voice-multihop radio network is considered. The network is represented as a connected graph, each node of the graph has a pre-specified number of transceivers. A voice-call uses one transceiver of each node that passes through, otherwise a call is blocked. Connected circuits, describing possible paths of calls, are pre-specified and calls arrive according to Poisson process. The goal is to minimize the blocking probability. The idea of the control policy is to optimize the number of calls in each circuit by imposing appropriate restrictions. Denoting \(x_j\) current number of calls in \(j\)-th circuit, the constrains can be written as \(x_j \leq X_j\), \(j=1,\dots,J\), \(\sum_{j \in S_l} x_j \leq Y_l\), \(l=1,\dots,k\), where each \(X_j\) is the threshold for \(j\)-th circuit and each \(Y_l\) is the threshold for the \(l\)-th group of circuits. The strategy can choose values \(X_j\), \(j=1,\dots,J\), groups \(S_l\), \(l=1,\dots,k\), and corresponding values \(Y_l\), \(l=1,\dots,k\). To find the optimal strategy the variant of simulated annealing algorithm with constant temperature is used. It accepts the hill climbing to escape the local solution traps. The standard clock simulation technique is used to estimate a performance of several policies simultaneously which are neighbors of the current policy. The simulation results are represented and indicate that the algorithm can locate an optimal or a near optimal solution quickly.