Download presentation
Presentation is loading. Please wait.
Published byEmery Joseph Modified over 9 years ago
1
16 Sep 04Transport Workshop at Queen's1 A numerical comparison of three heuristic methods for path reassignment for dynamic user equilibrium Ying-en Ge and Malachy Carey 16 September 2004 School of Management & Economics Queen’s University Belfast BT7 1NN
2
16 Sep 04Transport Workshop at Queen's2 Introduction Dynamic traffic assignment (DTA) 1.Network loading, with inflows/ assignment to spatial paths taken as given compute new path travel times 2.Spatial path reassignment (based on travel-times from 1) Three methods for path reassignment –Pair-wise swapping method –Wu et al. (1998) method –Lo & Szeto (2002) method
3
16 Sep 04Transport Workshop at Queen's3 Pair-wise swapping method Step 1 At iteration n, for each time interval i, note the path with current highest cost (travel time) and path with lowest cost [ or variants of this, e.g. choose the same paths for several time intervals, etc.] Step 2 For each time interval i, switch proportion s i n of inflow from higher cost to lower cost path s i n = n where n is a chosen parameter (1 > n > 0)
4
16 Sep 04Transport Workshop at Queen's4 Wu et al. (1998) method VI formulation The solution of the VI formulation is obtained by solving a series of quadratic programs below (1) where is a positive constant.
5
16 Sep 04Transport Workshop at Queen's5 Lo & Szeto (2002) method Step 1. Compute g ip n and v i n for all i and p by: g ip n = max{0, f ip n – [ ip n – ( u ip n – ( p f ip n – d i ))]} v i n = u i n - ( p g ip n – d i ) Step 2. Compute f ip n and u i n for all i and p f ip n+1 = f ip n - t n n (f ip n - g ip n ) u i n+1 = u i n - t n n ( u i n - v i n ) where t n = n (1 -0.25 -1 ), n (0,2) such that t n (0,1) and n = r 1 / r 2 with r 1 = ip ( f ip n - g ip n ) 2 + 2 i ( g ip n – g i ) 2 and r 2 = r 1 + i ( p f ip n – g ip n ) 2
6
16 Sep 04Transport Workshop at Queen's6 Numerical experiments Scenario Settings –2-link network –Network loading –Travel demand Convergence measure –Maximum absolute difference Numerical experiments –Effects of parameters in three methods –Convergence measure values over iterations, and –Accuracy of numerical solutions
7
16 Sep 04Transport Workshop at Queen's7
8
16 Sep 04Transport Workshop at Queen's8
9
16 Sep 04Transport Workshop at Queen's9 Minimum values of convergence measure Convergence measure Maximum absolute difference Pair-wise swapping method ( n = 1) 0.011116 Wu et al. method ( = 2) 0.058925 Lo & Szeto method ( = 0.5 and t n =1.0) 0.002874
10
16 Sep 04Transport Workshop at Queen's10
11
16 Sep 04Transport Workshop at Queen's11 Stopping iterations when given tolerances for maximum absolute difference are satisfied tolerance for maximum absolute difference 0.0625 (5%) 0.0125 (1%) 0.00625 (0.5%) Pair-wise swapping method ( n = 1) 17112 Wu et al. method ( = 2) 290 Lo & Szeto method ( = 0.50 and t n = 1.00) 305630783 Note: The percentages given in the round brackets after each tolerance represent the proportion of a tolerance to the free-flow travel time of the shorter of the two paths [1.25 minutes].
12
16 Sep 04Transport Workshop at Queen's12 Summary Preferred parameter values Not able to set an arbitrarily small tolerance Performance of three methods
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.