Membranealgorithms(MAs),whichinheritfromPsystems,constituteanewparallelanddistributeframeworkforapproximatecomputation.Inthepaper,amembranealgorithmisproposedwiththeimprovementthattheinvolvedparameterscanbeadaptivelychosen.Inthealgorithm,somemembranescanevolvedynamicallyduringthecomputingprocesstospecifythevaluesoftherequestedparameters.Thenewalgorithmistestedonawell-knowncombinatorialoptimizationproblem,thetravellingsalesmanproblem.Theempiricalevidencesuggeststhattheproposedapproachisefficientandreliablewhendealingwith11benchmarkinstances,particularlyobtainingthebestoftheknownsolutionsineightinstances.Comparedwiththegeneticalgorithm,simulatedannealingalgorithm,neuralnetworkandafine-tunednon-adaptivemembranealgorithm,ouralgorithmperformsbetterthanthem.Inpractice,todesigntheairlinenetworkthatminimizethetotalroutingcostontheCABdatawithtwenty-fiveUScities,wecanquicklyobtainhighqualitysolutionsusingouralgorithm.
