Correcting distorted demand distribution for overlapping paths

For every origin-destination relation the whole traffic demand will be distributed to all available paths. The distribution considers the general path costs, calculated from the measured variables and the configured weighting coefficients. A path consists of a sequence of edges. Two paths are different if their sequences of edges are not exactly the same. Two paths may also be considered to be different if they differ only by a small section. In such a case both paths would have about the same weight in the distribution of the traffic volume. This would lead to an overall biased distribution. This problem occurs in all assignment tasks and is called the blue/red bus paradox. This is depicted in the following figures:

Case 1: Two paths with identical cost

The distribution of trips 50:50 is unproblematic:

Case 2: Three paths with identical cost

The distribution of traffic on three paths is unproblematic. Each path receives one third of the demand:

Case 3: A slight variation results in 3 optional paths

Problem: Actually, there are only two quite dissimilar paths. Because of the slight variation in the end, the path search finds three different paths. Result: It is distributed amongst three paths. The overlapping part of the two similar paths receives too much traffic.

Case 4: Common stretches leads to three possible paths

The opposite of case 3: Actually there are 3 different paths but two of them have a small stretch in common. As in case 3, every path gets about one third of the demand. This is much more realistic compared to case 3.

Selecting the correction

You can correct the biased distribution of overlapping paths. Thus, the path selection model calculates a degree of commonality for the paths (commonality factor) The commonality factor expresses how much of a path is shared with other paths:

• Higher value: A path has many edges in common with other paths.
• Lower value: A path is largely independent from other paths.

Using this value the distribution function reduces the selection probability of paths with high commonality factor.

Notes: In certain network constellations, the correction of the biased distribution tends to spread traffic over longer paths if these paths have little in common with other paths. This can lead to unexpected results. In general, the correction of a biased distribution improves the result of the assignment. Use the correction of a biased distribution only in combination with restricting the cost difference between the paths.

1.  On the Traffic menu, click > Dynamic Assignment > Parameters.

The Dynamic Assignment: Parameters window opens.

2.  Select the Choice tab.

3.  In the Path choice model section, select Correction of overlapping paths.