Download presentation
Presentation is loading. Please wait.
1
Department of Industrial Engineering
Lecture 5: Algorithmic Methods for for finite Quasi-birth death processes Dr. Ahmad Al Hanbali Department of Industrial Engineering University of Twente Lecture 5: Finite QBDs
2
Lecture 5 This Lecture deals with continuous time Markov chains with finite state space and special structure as opposed to infinite space Markov chains in Lecture 3 and 4 Objective: To find equilibrium distribution of the Markov chain Lecture 5: Finite QBDs
3
Finite Quasi-Birth Death processes
In many applications, the level is the number of customers in a system can be finite Subset of state space with common ๐ entry is called level ๐ (0โค ๐โค๐) and denoted ๐(๐)={(๐,0),(๐,1),โฆ,(๐,๐โ1)}. This means state space is restricted โช 0โค๐โค๐ ๐(๐) The generator of the irreducible continuous time finite QBD has the following form ๐= ๐ต 00 ๐ด 2 0 โฎ โฎ โฎ 0 ๐ด 0 ๐ด 1 ๐ด 2 โฑ โฑ โฑ โฑ 0 ๐ด 0 ๐ด 1 โฑ โฑ โฑ โฑ ๐ด 0 โฑ โฑ ๐ด โฆ โฑ โฑ โฑ โฑ ๐ด 1 ๐ด โฑ โฑ โฑ โฑ ๐ด 0 ๐ต ๐๐ How to find the equilibrium probabilities, ๐๐=0? Three methods Lecture 5: Finite QBDs
4
Method 1: Linear level reduction
Let us define the following matrices: ๐ถ 0 = ๐ต 00 , ๐ถ ๐ = ๐ด 1 โ ๐ด 2 ๐ถ ๐โ1 โ1 ๐ด 0 , 1โค๐โค๐โ1, ๐ถ ๐ = ๐ต ๐๐ โ ๐ด 2 ๐ถ ๐โ1 โ1 ๐ด 0 , - ๐ถ ๐ โ1 ๐ด 0 records first passage probabilities from ๐ ๐ to ๐ ๐+1 Theorem: the equilibrium probability ๐= ๐ 0 ,โฆ, ๐ ๐ is determined by: ๐ ๐ ๐ถ ๐ =0, ๐ ๐ =โ ๐ ๐+1 ๐ด 2 ๐ถ ๐ โ1 , โค๐โค๐โ1, ๐=0 ๐ ๐ ๐ ๐ =1 In the theorem here we are cutting off the levels from the level 0 and moving up to level M. One might also proceed in the reverse direction by starting to cut off level M down to level 0. Lecture 5: Finite QBDs
5
Method 2: Method of Folding (1)
Assume that ๐=2K. Partition the state space into two subsets ๐ธ with even numbered levels and ๐ธ ๐ with odd numbered levels Reorder the levels of finite QBD such that the levels in ๐ธ comes first. Then Q becomes: ๐= ๐ต ๐ด ๐ด ๐ด 0 ๐ด โฑ โฑ โฑ ๐ด ๐ด 0 ๐ด ๐ต ๐๐ ๐ด 0 ๐ด 0 ๐ด ๐ด ๐ด 0 ๐ด ๐ด โฑ โฑ โฑ ๐ด 0 ๐ด ๐ด 1 , This gives that: ๐ 2๐+1 =โ ๐ 2๐ ๐ด 0 + ๐ 2๐+2 ๐ด 2 ๐ด 1 โ1 , ๐=0,โฆ,๐พโ1 Lecture 5: Finite QBDs
6
Method 2: Method of Folding (2)
The vector ๐ 0 , ๐ 2 ,โฆ, ๐ 2๐พ is proportional to the equilibrium probability ๐ 0 โ , ๐ 1 โ ,โฆ, ๐ ๐พ โ vector of the chain restricted to even numbered levels with generator ๐ โ = ๐ต 00 โ ๐ด 2 โ 0 โฎ โฎ โฎ โฎ ๐ด 0 โ ๐ด 1 โ ๐ด 2 โ โฑ โฑ โฑ โฑ 0 ๐ด 0 โ ๐ด 1 โ โฑ โฑ โฑ โฑ ๐ด 0 โ โฑ โฑ ๐ด 2 โ โฑ โฆ โฑ โฑ โฑ โฑ ๐ด 1 โ ๐ด 2 โ โฆ โฑ โฑ โฑ โฑ ๐ด 0 โ ๐ต ๐๐ โ , ๐ต 00 โ = ๐ต 00 โ A 0 A 1 โ1 A 2 , ๐ด 0 โ =โ ๐ด 0 ๐ด 1 โ1 ๐ด 0 , ๐ด 2 โ =โ ๐ด 2 ๐ด 1 โ1 ๐ด 2 , ๐ด 1 โ = ๐ด 1 โ ๐ด 2 ๐ด 1 โ1 ๐ด 0 โ ๐ด 0 ๐ด 1 โ1 ๐ด 2 , ๐ต ๐๐ โ = ๐ต ๐๐ โ A 2 A 1 โ1 A 0 To solve a QBD with ๐ levels it suffices to solve QBD with ๐/2 levels. Repeating folding on smaller QBD we obtain QBD with ๐/4 levels, and so forth until 2 levels Lecture 5: Finite QBDs
7
Method 3: Matrix geometric combination
Let ๐
be the minimal nonnegative solution of ๐ด 0 +๐
๐ด 1 + ๐
2 ๐ด 2 =0 Let ๐
be the minimal nonnegative solution of ๐ด 2 + ๐
๐ด 1 + ๐
2 ๐ด 0 =0 Theorem Let ๐ด= ๐ด 0 + ๐ด 1 + ๐ด 0 be irreducible and ๐๐ด= 0, ๐๐=1. If ๐ ๐ด 0 ๐โ ๐ ๐ด 2 ๐, the equilibrium probability of the finite QBD is given by ๐ ๐ = ๐ฅ 0 ๐
๐ + ๐ฅ ๐ ๐
๐โ๐ , ๐=0,โฆ,๐ where ๐ฅ 0 , ๐ฅ ๐ is the solution of the system ๐ฅ 0 , ๐ฅ ๐ ๐ต 00 +๐
๐ด 2 ๐
๐ ๐ต ๐
๐โ1 ๐ด 2 ๐
๐ ๐ต ๐๐ + ๐
๐โ1 ๐ด 0 ๐ต ๐๐ + ๐
๐ด 0 =0 with ๐ฅ 0 ๐=0 ๐ ๐
๐ ๐+ ๐ฅ ๐ ๐=0 ๐ ๐
๐ ๐=1 Let ๐
be the minimal nonnegative solution of (level reversed process) Lecture 5: Finite QBDs
8
Example: uninterrupted traffic on a highway
Level dependent and independent QBDs were applied to mimic the traffic behavior on highways especially the fundamental diagram (flow-density diagram) To model this Niek Baer used the so-called four stage M/M/1 threshold queues Lecture 5: Finite QBDs
9
References Niek Baer. Queueing and Traffic, PhD thesis University of Twente 2015 G. Latouche and V. Ramaswami (1999), Introduction to Matrix Analytic Methods in Stochastic Modeling. SIAM. Lecture 5: Finite QBDs
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.