Affiliations: Department of Electronic Engineering, National Taipei University of Technology, Taipei, Taiwan, ROC Tel.: +886 2 2771 2171, Ext 2275/2200; Fax: +886 2 2731 7120; E‐mail: [email protected] | EEE, Hong Kong UST, Hong Kong, PR China E‐mail: [email protected]
Note: [] Corresponding author. Mailing address: No. 1, Sec. 3, Chung‐Hsiao E. Rd., Taipei 10643, Taiwan, ROC.
Note: [] Parts of this paper were presented at the IEEE Conference on Communications (ICC ’96), Dallas, TX, June 23–27, 1996.
Abstract: The large number of states that arises in a broadband network makes the design and analysis of call admission policies in a broadband network computationally intractable. State reduction, therefore, is an important step to tackle many broadband networking issues. In this paper we propose a method to reduce the state space. An important characteristic of the method is that the total gain of a policy remains unchanged in the reduced state space: If a policy improves performance in the reduced space, it does likewise in the original space. This means that the search for a better policy can be conducted in the reduced state space created by the state reduction method introduced in this paper. We present several examples to demonstrate the applications of the method for state reduction.
Keywords: Multirate circuit‐switched networks, coordinate convex policy, state grouping, Markov decision process
