Cc. Lo et Wh. Chang, A multiobjective hybrid genetic algorithm for the capacitated multipoint network design problem, IEEE SYST B, 30(3), 2000, pp. 461-470
Citations number
36
Categorie Soggetti
AI Robotics and Automatic Control
Journal title
IEEE TRANSACTIONS ON SYSTEMS MAN AND CYBERNETICS PART B-CYBERNETICS
The capacitated multipoint network design problem (CMNDP) is NP-complete, I
n this paper, a hybrid genetic algorithm for CMNDP is proposed. The multiob
jective hybrid genetic algorithm (MOHGA) differs from other genetic algorit
hms (GA's) mainly in its selection procedure. The concept of subpopulation
is used in MOHGA, Four subpopulations are generated according to the elitis
m reservation strategy, the shifting Prufer vector, the stochastic universa
l sampling, and the complete random method, respectively, Mixing these four
subpopulations produces the next generation population. The MOHGA can effe
ctively search the feasible solution space due to population diversity, The
MOHGA has been applied to CMNDP. By examining computational and analytical
results, we notice that the MOHGA can find most nondominated solutions and
is much more effective and efficient than other multiobjective GA's.