Adaptive multi-objective optimization based on feedback design

Abstract

The problem of adaptive multi-objective optimization (AMOO) has received extensive attention due to its practical significance. An important issue in optimizing a multi-objective system is adjusting the weighting coefficients of multiple objectives so as to keep track of various conditions. In this paper, a feedback structure for AMOO is designed. Moreover, the reinforcement learning combined with hidden biasing information is applied to online tuning weighting coefficients of objective functions. Finally, the proposed approach is applied to the optimization design problem of an elevator group control system. Simulation results show that AMOO has the best average performance at up-peak traffic profile, and its average waiting time reaches 22 s. AMOO is suitable for various traffic patterns, and it is also superior to the majority of algorithms at down-peak traffic profile.

Keywords

multi-objective optimization / adaptive optimization / reinforcement learning / elevator group system

References

[1] Fujino A., Tobita T., Segawa K., et al. An elevator group control system with floor attribute control method and system optimization using genetic algorithms[J]. IEEE Transactions on Industrial Electronics, 1997, 44(4): 546-552.

[2] Li S. Y., Hu C. F. An interactive satisfying method based on alternative tolerance for multiple objective optimization with fuzzy parameters[J]. IEEE Transactions on Fuzzy Systems, 2008, 16(5): 1151-1160.

[3] Konak A., Coit D. W., Smith A. E. Multi-objective optimization using genetic algorithms: A tutorial[J]. Reliability Engineering and System Safety, 2006, 91(9): 992-1007.

[4] Hirasawa K., Eguchi T., Zhou J., et al. A double-deck elevator group supervisory control system using genetic network programming[J]. IEEE Transactions on Systems, Man, and Cybernetics, Part C: Applications and Reviews, 2008, 38(4): 535-550.

[5] Sung K. O., Witold P. Genetic optimization driven multilayer hybrid fuzzy neural networks[J]. Simulation Modelling Practice and Theory, 2006, 14(5): 597-613.

[6] Kim C. B., Seong K. A., Kwang H. L. Design and implementation of a fuzzy elevator group control system[J]. IEEE Transactions on Systems, Man and Cybernetics, Part A: Systems and Humans, 1998, 28(3): 277-287.

[7] Andrea B., Lazaric A., Francesco M., et al. Reinforcement distribution in fuzzy Q-learning[J]. Fuzzy Sets and Systems, 2009, 160(10): 1420-1443.

[8] Ma X., Likharev K. K. Global reinforcement learning in neural networks[J]. IEEE Transactions on Neural Networks, 2007, 18(2): 573-577.

[9] Millan R., Posenato D., Dedieu E. Continuous-action Q-learning[J]. Machine Learning, 2002, 49(2): 247-265.

[10] Zou X. F., Chen Y., Liu M. Z., et al. A new evolutionary algorithm for solving many-objective optimization problems[ J] IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics, 2008, 38(5): 1402-1412.

[11] Sutton R. S., Barto A. G. Reinforcement Learning: An Introduction[M]. 1998, Cambridge, MA: MIT Press.

[12] Kamal M. A. S., Murata J. Reinforcement learning for problems with symmetrical restricted states[J]. Robotics and Autonomous Systems, 2008, 56(9): 717-727.

[13] Skelly M. M. Hierarchical Reinforcement Learning with Function Approximation for Adaptive Control[D]. 2004, Cleveland, USA: Case Western Reserve University.

[14] Cortes P., Larraneta J. A Genetic Algorithm for Controlling Elevator Group Systems[M]. 2003, Berlin: Springer.

[15] Zong Qun, Dou Liqian, Wang Weijia. Elevator group control scheduling approach based on multi-agent coordination[C]. In: The 6th World Congress on Intelligent Control and Automation. Dalian, China, 2006. 7249–7253.