Tools/Turns

jtcrouter.py#

The JunctionTurnCountRouter generates vehicle routes from turn-count data. It does so by converting the turn counts into into flows and turn-ratio files that are suitable as jtrrouter input. Then it calls jtrrouter in the background. The turn

<SUMO_HOME>/tools/jtcrouter.py -n <net-file> -t <turn-file> -o <output-file>

There are three basic styles of converting turn-counts to routes:

  • Flows start at all turn-count locations in the network but end when reaching the next count location
  • Flows start at all turn-count locations in the network and are discounted when reaching the next count location (--discount-sources)
  • Flows only start on the fringe of the network (--fringe-flows)

Turn count data format#

The turn-count data must be provided in the format:

<interval id="generated" begin="0.0" end="99.0">
        <edgeRelation from="-58.121.42" to="64" count="1"/>
        <edgeRelation from="-58.121.42" to="-31" count="3"/>
        <edgeRelation from="45" to="-68" count="3"/>
        <edgeRelation from="-31.80.00" to="31" count="1"/>
        <edgeRelation from="-31.80.00" to="37" count="1"/>
        <edgeRelation from="-31.80.00" to="-23" count="13"/>
        <edgeRelation from="-92.180.00" to="-60" count="1"/>
</interval>

routeSampler.py#

The script generates routes from turn-count data, edge-count and even origin-destination-count data. It requires a route file as input that defines possible routes. Routes are sampled (heuristically) from the input so that the resulting routes fulfill the counting data

turn counts#

The turn-count data format is the same as as described above.

<SUMO_HOME>/tools/routeSampler.py -r <input-route-file> --turn-files <turn-files -o <output-file>

The attributes for reading the counts from the turn-data file can be set with option --turn-attribute (default 'count')

edge counts#

In addition to loading a turn-count file, routeSampler can also load an edgeData file using option --edgedata-files.

<SUMO_HOME>/tools/routeSampler.py -r <input-route-file> --edgedata-files <edgedata-files -o <output-file>

The attributes for reading the counts from edge-data file can be set with option --edgedata-attribute (default 'entered')

Note

When using routeSampler as a replacement for dfrouter or flowrouter.py, the flow input files can be converted to edgeData files with the tool edgeDataFromFlow.py

Obtaining initial routes#

Routes generated by randomTrips.py (--route-output) can be a good input source. The following randomTrips.py options may be helpful:

  • --fringe-factor: setting a high value will generated lots of through-traffic which is plausible for small networks
  • --min-distance: restricting short routes increases the chance that routes passing multiple counting locations are generated
  • --speed-exponent, --lanes: both options can be used to increase routes starting and ending on "important" roads. This combines well with the --fringe-factor-option to generate many routes that enter and leave the network on major roads.

Example:

<SUMO_HOME>/tools/randomTrips.py -n <input-net-file> -r sampleRoutes.rou.xml
<SUMO_HOME>/tools/routeSampler.py -r sampleRoutes.rou.xml --edgedata-files <edgedata-files -o <output-file>

Generalized route restrictions#

By default, the input options --edgedata-files and --turn-files allow restricting counts for single edges and pairs of consecutive edges.

To define count restrictions for longer consecutive edge sequences the optional 'via' attribute can be used for <edgeRelation> elements:

   <edgeRelation from="A" to="D" via="B C" count="42"/>

To define count restrictions on non-consecutive edges the option --turn-max-gap can be used. Example: When setting option --turn-max-gap 2, the edgeRelation <edgeRelation from="A" to="D" .../> would apply to routes containing "A B", "A X D" or "A X Y D" but not "A X Y Z D".

Origin-Destination restrictions#

When loading an edgeRelation file with the option --od-files, origin-destination counts will be added. This can be used to combine (edge-based) OD-relations with other counting data.

Output Styles#

By default, routeSampler will generate individual vehicles with embedded routes. This can be changed with the following options (which can also be combined):

  • --write-route-ids: write named routes and let vehicles reference the route via it's id
  • --wroute-route-distribution STR: put all routes into a route distribution (with appropriate route probabilities) and let a all vehicles reference this distribution (Simulation counts will then vary due to sampling from the distribution)
  • --write-flows number: write <flow>-definitions instead of vehicles. The exact number of flow vehicles will be spaced evenly between the earliest and latest vehicle departures that would have been generated by default
  • --write-flows probability: write <flow>-definitions instead of vehicles. Flows will be defined with attribute 'probability' so that the expected number of vehicles is equal to the number of vehicles that would have been generated by default but the specific number will vary due to sampling effects

Vehicle attributes#

With the option --attributes <STRING>, additional parameters can be given to the generated vehicles (note, usage of the quoting characters).

<SUMO_HOME>/tools/randomTrips.py -n input_net.net.xml 
  --attributes="departLane=\"best\" departSpeed=\"max\" departPos=\"random\""

This would make the rvehicles be distributed randomly on their starting edges and inserted with high speed on a reasonable lane.

Caution

Quoting of trip attributes on Linux must use the style --attributes 'departLane="best" departSpeed="max" departPos="random"'

Sampling#

By default, sampling will be performed iteratively by 1) selecting a random counting location that has not yet reached it's count (and which still has viable routes) 2) selecting a random route that passes this counting location

until all counting locations have reached their measured count or there are no viable routes (routes which have all their passed counting locations below the input count)

By setting the option --weighted. The sampling algorithm is changed. For each route a probability value is loaded from the input. The probabilty can either be specified explicitly using route attribute 'probability' or implicitly if a route with the same sequence of edges appears multiple times in the the route input. Sampling will be performed iteratively by 1) selecting a random viable route sampled by probability

until all counting locations have reached their measured count or there are no viable routes (routes which have all their passed counting locations below the input count)

Optimization#

By default, routes will be sampled from the input route set until no further routes can be added without exceeding one of the counts. This may still leave some counts below their target values. At this point an ILP-Solver can be used to swap out routes and get closer to the target values or even reach the exact numbers. By setting option --optimize . The number of times that a route is used can be changed by up to times. This defines a trade-off between using routes in the same distribution as found in the input and optimizing the counts. When setting option --optimize full. No constraints on the route distribution are set and any route can be used as often as needed to reach the counts.

Further Calibration#

It is possible to load the resulting output into routeSampler.py for another round of optimization. By setting the option --optimize-input the sampling step is skipped and the optimizer is run directly on the input route set.

By removing specific routes or adding new ones, the user can thus tailor the generating traffic in an iterative manner.

generateTurnRatios.py#

This script is used to calculate the turn ratios from a an edge to its downstream edge with a given route file. The output file can be directly used as input in jtrrouter. The time interval will span the minimum and maximum departure times of the route file.

<SUMO_HOME>/tools/generateTurnRatios.py -r <route-file>

The standard output is the traffic volumes (which jtrrouter normalizes automatically). With the option -p, turning ratios will be written as values from [0,1].

generateTurnDefs.py#

This script allows generation of the turn definitions based on the number of lanes allowing particular turns. The basic functionality distributes the traffic uniformly, that is:

  1. distribute the incoming traffic uniformly across the lanes forming the road
  2. distribute the amount of traffic assigned to each lane uniformly among the destinations that the lane allows turns to.
  3. sum up the values for each of the destinations that the road being processed allows turning to.

Example use

<SUMO_HOME>/tools/turn-defs/generateTurnDefs.py --connections-file connections.con.xml --turn-definitions-file output.turndefs.xml

The script allows to be extended with new traffic distribution policies (for example, based on Gaussian distribution) easily. See the DestinationWeightCalculator class for details.

The script processes the connections given in the provided *.con.xml file. For usage details, execute the generateTurnDefs.py script with --help option.

Note

You can generate a connections file with all the connections in the network using netconvert - see the --plain-output-prefix option.

turnCount2EdgeCount.py#

This script converts turn-count data into edgeData.

<SUMO_HOME>/tools/turn-defs/turnCount2EdgeCount.py -t <turn-file> -o <output-file>

turnFile2EdgeRelations.py#

This script converts the deprecated turn-file format into edgeRelation format

<SUMO_HOME>/tools/turn-defs/turnFile2EdgeRelations.py -t <turn-file> -o <output-file>