The new DRT estimate and teleport module (a continuation based on PR #3160) will generate a trip estimation (estimated waiting time, travel time and the probability of being served) for each DRT request. The DRT trips will then be teleported during the MobSim. A more detailed documentation will follow shortly. Some reformatting will be performed before this PR is merged into master branch. Ideas and comments are welcomed!!!
A brief intro to this implementation is provided here:
There are several advantages of doing this:
Speed up the procedure
Enhancing the reliability of the results
Easier for transport planning
Being able to match real-world operational data
Current implementation
Interface for typical waiting time estimator
a. Constant waiting time estimator
b. Shape-file based waiting time estimator
Interface for typical ride duration estimator
a. Network based ride duration estimator
b. Euclidean distance based ride duration estimator
Distribution generator
a. Log-normal distribution generator
b. Normal distribution generator
c. No distribution (currently can be set via normal distribution with std = 0)
Probability of being accepted
a. Uniform value
b. shape-file-based (to-do)
How to use:
To build a DRT estimator, we need to choose a waiting time estimator (1) and a ride duration estimator (2). For each of the two estimators, we can match it with an independent distribution model (3). Finally, we choose an estimation model for the probability of being accepted.
The parameters for the estimators and the distribution models can be fitted according to real-world operational data or systematic offline DRT simulations (i.e., based on fix demands). Other simulation tools or optimizers can also be used for this purpose.
The new DRT estimate and teleport module (a continuation based on PR #3160) will generate a trip estimation (estimated waiting time, travel time and the probability of being served) for each DRT request. The DRT trips will then be teleported during the MobSim. A more detailed documentation will follow shortly. Some reformatting will be performed before this PR is merged into master branch. Ideas and comments are welcomed!!!
A brief intro to this implementation is provided here: There are several advantages of doing this:
Current implementation
Interface for typical waiting time estimator a. Constant waiting time estimator b. Shape-file based waiting time estimator
Interface for typical ride duration estimator a. Network based ride duration estimator b. Euclidean distance based ride duration estimator
Distribution generator a. Log-normal distribution generator b. Normal distribution generator c. No distribution (currently can be set via normal distribution with std = 0)
Probability of being accepted a. Uniform value b. shape-file-based (to-do)
How to use:
To build a DRT estimator, we need to choose a waiting time estimator (1) and a ride duration estimator (2). For each of the two estimators, we can match it with an independent distribution model (3). Finally, we choose an estimation model for the probability of being accepted.
The parameters for the estimators and the distribution models can be fitted according to real-world operational data or systematic offline DRT simulations (i.e., based on fix demands). Other simulation tools or optimizers can also be used for this purpose.