This page describes simulations of trains in SUMO. To build an intermodal simulation scenario with trains, additional steps have to be taken in comparison to a plain vehicular simulation.
Building a network for train simulation#
Railways can be imported from OSM. They can also be explicitly specified using the existing vClasses.
If a railway has parallel tracks, these must be modelled as distinct edge elements instead of a single edge with multiple "lanes".
When importing from OSM the following railway types are distinguished by default (via <SUMO_HOME>/data/typemap/osmNetconvert.typ.xml):
Whenever a railway is electrified, the vClasses rail, rail_electric and rail_fast are permitted to drive there. Otherwise, only vClass rail is allowed.
By loading the additional typemap <SUMO_HOME>/data/typemap/osmNetconvertRailUsage.typ.xml, additional usage information is exported for the edge types:
This will lead to compound edge type ids such as railway.rail|usage.main.
Local track numbers (mostly in train stations) are exported as edge parameter track_ref. These values are shown in the edge parameter dialog and can also be used for coloring (color by param, streetwise).
Bidirectional track usage#
In reality all tracks can be used in either direction if the need arises but many rails are used in only one direction most of the time. In SUMO, bidirectional track usage must be enabled explicitly. This simplifies routing as rails will only be used in their preferred direction most of the time.
Bidirectional track usage is modeled by two edges that have their
geometries exactly reversed and using the attribute
spreadType="center". This will result
in lane geometries that are overlaid exactly. These edges are referred
to as superposed (alternatively as bidirectional rail edges). In the
.net.xml file these edges are marked with
bidi="<REVERSE_EDGE_ID>" but this is a generated
attribute and not to be set by the user.
When Rail signals are placed at both ends of a bidirectional track they will restrict it's usage to one direction at a time.
Bidirectional rails in sumo-gui#
Visualization of bidirectional tracks has a distinct style and dedicated settings.
Working with bidirectional tracks in netedit#
- To show both edges that constitute a bidirectional track, activate
edge visualization option spread bidirectional railways/roads. Both edges will be
drawn narrower and with a side-offset to make them both visible
- this is highly recommended when using connection mode to define connections among bidirectional tracks as it's otherwise hard to distinguish the affected edges
- the pre-defined gui setting scheme 'rail' automatically activates the spread ... setting.
- To find (and highlight) all bidirectional tracks, use attribute selection and search for attribute bidi with a value of 1
- Create bidirectional tracks as explained here
Routing in bidirectional networks#
When computing train routes in a network with parallel tracks which are usable in both directions, it may be desirable that trains preferentially use one of the tracks (i.e. to always keep on the right side) and thereby preventing conflicts between oncoming trains.
To express this preference, the edges in the preferred direction and on the preferred side may be assigned a higher priority value. This value will be taken into account when setting option --weights.priority-factor which applies to sumo and duarouter.
Importing bidirectional tracks from OSM#
When importing networks from
OSM, rails tagged with
automatically imported as superposed edges.
Handling Problems in bidirectional railway networks#
Commonly, rail networks import from OSM are incomplete in regard to bidirectional track usage. One example would be terminal tracks which a train can only leave by reversing direction. netconvert provides several options to deal with this problem:
- --railway.topology.output <FILE>: Identify problematic tracks (i.e. for manual correction).
- --railway.topology.repair : Automatically fixes problems by analyzing rail network topology and adding bidirectional tracks
- --railway.topology.repair.minimal : Apply automated fixes but only those which are needed for the operation loaded public transport lines
- --railway.topology.repair.connect-straight : Allow bidirectional rail use wherever rails with opposite directions meet at a straight angle
- --railway.topology.repair.stop-turn : Add turn-around connections at all loaded stops (to enable direction reversal).
- --railway.topology.all-bidi : make all tracks usable in both directions.
- --railway.topology.all-bidi.input-file : Load edge ids from text file and make those tracks usable in both directions. (one edge id per line, optionally prefixed with 'edge:' as in a selection file created with netedit).
The node type
rail_signal may be used to define signals which implement Automatic Block
By setting the netconvert-option --railway.signals.discard all signals can be removed from a network.
When working with bidirectional tracks, rail signals will affect both directions of travel by default. This can be changed in netedit.
The node type
rail_crossing may be used to define railway crossings. At these nodes trains will
always have the right of way and road vehicles get a red light until
there is a safe gap between approaching trains.
When importing networks from OpenStreetMap, rail crossings will be imported automatically. For other input data sources the crossings may have to be specified via additional xml files or set via netedit after importing.
Rail Crossing Parameters#
The following parameters may be defined in an additional file to configure timing and behavior of a rail crossing (example below). Time values are in seconds.
- time-gap : time headway of approaching train that triggers closing the crossing (default 15)
- space-gap : space headway of approaching train that triggers closing the crossing (default -1, which disables this check)
- min-green-time : minimum green time for opening the crossing when another train is approaching (default 5)
- opening-delay : time after train has passed the crossing before the opening sequence starts (default 3)
- opening-time : time for opening the crossing (indicated by red-yellow state 'u') (default 3)
- yellow-time : time for closing the crossing (default 3)
When setting parameters,
id indicates the id of the rail-crossing junction and
programID is always '0'.
<additional> <tlLogic id="C" programID="0"> <param key="time-gap" value="15.0"/> <param key="space-gap" value="-1.0"/> <param key="min-green-time" value="5.0"/> <param key="opening-delay" value="3.0"/> <param key="opening-time" value="3.0"/> <param key="yellow-time" value="5.0"/> </tlLogic> </additional>
Kilometrage (Mileage, Chainage)#
Edges support the attribute distance to denote the distance at the start of the edge relative to some point of reference for a linear referencing scheme. When the distance metric decreases along the forward direction of the edge, this is indicated by using a negative sign for the distance value.
The distance value along an edge is computed as:
|edgeDistance + vehiclePos|
Edge distance is imported from OSM and can also be be set along a route in netedit
The distances value can be written in fcd-output using option --fcd-output.distance. It may then be used for plotting by plot_trajectories.py using the code
k (i.e. -t kt). The distances can also be visualized in sumo-gui (color edges by distance).
Negative distance values are not currently supported (pending introduction of another attribute)
Routing on Bidirectional Tracks#
When train tracks can be used in both directions, there is considerable freedom for trains when search a path through the network. To reduce the number of conflicts (when two vehicles want to use the same track in opposite directions), the preferred direction for each track can be defined and factored into the routing decision.
When routes are computed in the simulation, this is done by setting the option --device.rerouting.priority-factor FLOAT. This causes the priority values of edges to be factored into the routing decision with higher values being preferred. At the default value of 0. Edge priority is ignored when routing. When setting a positive value, the edges with the lowest priority receive a penalty factor to their estimated travel time of 1 + FLOAT whereas the edges with the highest priority receive no penalty. Edges with medium priority will receive a penalty of 1 + x * FLOAT where
x = (edgePriority - minPriority) / (maxPriority - minPriority)
The priority values can either be assigned by the user or computed heuristically by netconvert by setting the option --railway.topology.direction-priority. This requires that some of the tracks in the network are uni-directional (to unambiguously define the main direction). The assigned priority values are:
- 4: unidirectional track
- 3: main direction of bidirectional track
- 2: undetermined main direction (straight continuation from different directions of unidirectional track)
- 1: undetermined main direction (no continuation from unidirectional track)
- 0: reverse of main direction of bidirectional track
There is a dedicated carFollowMode for trains which can be activated by
carFollowModel="Rail" trainType="<TYPE>" in the
<vType> definition. Allowed values for trainType are
These types model traction and rolling resistance for particular trains. Alternatively, any other car following model may be used and configured with appropriate acceleration / deceleration parameters.
Train schedules are defined in the same way as for any other type of public transport.
Generated by plotXMLAttrributes.py.
When simulation trains on a network with railway signals, trains will
only enter a block (a section of edges between signals) if it is free of
other trains. When there are no rail signals or multiple trains have
been inserted in the same block, they will automatically keep a safe
distance according to their car following model. When using
will always keep enough distance to the leading train to come to a safe
stop even if the lead train was to stop instantly.
Trains will reverse direction if all of the following conditions are met:
- The head of the train is on a normal edge (not on an intersection / railway switch)
- The whole length of the train is located on rail-edges that allow bidirectional use.
- The speed of the train is below 0.1m/s.
- The train does not have any further stops on the current edge
- The succeeding edges in the train's route are the reverse-direction edges of those it is currently on
There is a "turn-around" connection from the current train edge to the reverse direction edge
When importing public transport stops with option --ptstop-output, all bidirectional edges with a public transport stop will have the necessary turn-around connection and thus be eligible for reversing.
Trains can be split and joined (divided and coupled) at stops.
Splitting a train#
To split a train, the following input definition can be used. The rear half of the train is defined as a new vehicle which depart value split. The train train that is being split must define the 'split' attribute in its stop definition referencing the id of the rear half.
<vType id="train" vClass="rail"/> <vType id="splitTrain" vClass="rail" length="50"/> <trip id="t0" type="train" depart="0.00" from="a" to="c"> <stop busStop="B" duration="60.00" split="t1"/> </trip> <trip id="t1" type="splitTrain" depart="split" departPos="last" from="b" to="e"> <stop busStop="B" duration="60.00"/> </trip>
When defined this way, The rear part of the train will be created as a new simulation vehicle once the first part has reached the stop. After stopping, The front half of the train will continue with reduced length.
Joining two trains#
To join two trains, the following input definition can be used. The front half of the train must define a stop trigger with value join. The rear half of the other train must define the attribute 'join' referencing the id of the front half.
<vType id="train" vClass="rail"/> <vType id="splitTrain" vClass="rail" length="50"/> <trip id="t0" type="splitTrain" depart="0.00" from="a" to="c"> <stop busStop="B" duration="60.00" triggered="join"/> </trip> <trip id="t1" type="splitTrain" depart="30" from="d" to="b"> <stop busStop="B" duration="5.00" join="t0"/> </trip>
The rear part of the train will be joined to the front part if the following conditions are met: - the rear part has fulfilled its stopping duration - the front part the train is present and it's back is on the same lane as the front of the rear part - the gap between the trains is less than 5 meters After being joined to the front part, the rear part will no longer be part of the simulation. The front half of the train will stop until the rear part is joined to it. Afterwards it will continue with increased length.
Rail Signal Behavior#
Rail signals perform the following safety functions automatically
- a) guard the track up to the next rail signal (signal block) so that only one train can enter this section at a time. This prevents rear-end collisions.
- b) guard the track so that vehicles from different branches (flanks) cannot enter the same section. This prevents flanking collisions.
- c) guard the track so that vehicles cannot enter bidirectional sections at the same time. This prevents head-on collisions.
- d) prevent deadlocks on bidirectional sections
Functionality a) corresponds to the "classic" safety behavior of rail signals (PZB). When option --railsignal-moving-block is set or individual signals are configured with parameter moving-block (see below), feature a) is disabled and trains will use their configured carFollowModel (i.e. 'Rail') for distance keeping. This is similar to the LZB safety system when used with extremely short virtual blocks.
To switch a single signal into moving-block-mode, the following additional file may be loaded:
<additional xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:noNamespaceSchemaLocation="https://sumo.dlr.de/xsd/additional_file.xsd"> <tlLogic id="gneJ8" programID="0"> <param key="moving-block" value="true"/> </tlLogic> </additional>
Parameter moving-block may also be updated at runtime with
Additionally, rail signals can enforce train ordering to ensure that a scheduled order at stations can be kept. To make use of this, the following elements can be loaded from an additional file:
<railSignalConstraints id="A"> <predecessor tripId="t0" tl="D" foes="t1" limit="2"/> <predecessor tripId="t0" tl="C" foes="t2"/> <insertionPredecessor tripId="t3" tl="E" foes="t4"/> </railSignalConstraints>
This constrain defines that a given vehicle id (or tripId) can only pass the current signal after some other vehicle ('foe') with the given id or tripId has passed signal 'tl'. The foe vehicle must have been the last vehicle to do so or it must have been one of the last 'limit' vehicles at the time of switching green.
This constrain defines that a given vehicle id (or tripId) can only be inserted on the block leading up to the current signal after some other vehicle ('foe') with the given id or tripId has passed signal 'tl'. The foe vehicle must have been the last vehicle to do so or it must have been one of the last 'limit' vehicles at the time of switching green.
|Attribute Name||Value Type||Description|
|tripId||id (string)||The 'tripId' param of a vehicle or the id of a vehicle to which this constraint applies|
|tl||id (string)||The id of a railSignal|
|foes||ids (string list)||The ids of one or more vehicles that must have passed tl before the parent rail signal permits tripID to pass|
|limit||int||The number of intermediate vehicles that may pass tl after the foes before the consraint is evaluated for tripId. default: number of given foes. (setting a high number has now downside besides memory use but setting a low number may cause the constraint to block tripId indefinitely because the ids of the passed foes were overwritten by later trains)|
|active||bool||Whether this constraint is active (inactive constraints may still be retrieved via TraCI)|
Constraints can be generated using the tool generateRailSignalConstraints.py by using a route file with stops that define a schedule.
Rail signals and rail crossings can be controlled with function traci.trafficlight.setRedYellowGreenState. They can also be switched off with traci.trafficlight.setProgram(tlsID, "off"). In either case, normal operations can be resumed by reactivating the default program "0": traci.trafficlight.setProgram(tlsID, "0").
Trains can be controlled just like cars by using the traci.vehicle functions. Furthermore the following functions are available for rail signals:
- traci.trafficlight.getBlockingVehicles(tlsID, linkIndex): Returns the list of vehicles that are blocking the subsequent block for the given tls-linkIndex from the perspective of the closest vehicle upstream of the signal
- traci.trafficlight.getRivalVehicles(tlsID, linkIndex): Returns the list of vehicles that also wish to enter the subsequent block for the given tls-linkIndex (regardless of priority) from the perspective of the closest vehicle upstream of the signal
- traci.trafficlight.getPriorityVehicles(tlsID, linkIndex): Returns the list of vehicles that also wish to enter the subsequent block for the given tls-linkIndex (only those with higher priority) from the perspective of the closest vehicle upstream of the signal
Constraints can be queried and modified via TraCI:
- getConstraints(tlsID, tripId=""): Returns the list of rail signal constraints for the given rail signal. If tripId is not "", only constraints with the given tripId are returned. Otherwise, all constraints are returned
- getConstraintsByFoe(foeSignal, foeId=""): Returns the list of rail signal constraints that have the given rail signal id as their foeSignal. If foeId is not "", only constraints with the given foeId are returned. Otherwise, all constraints are returned
- swapConstraints(tlsID, tripId, foeSignal, foeId): Reverse the given constraint and return list of new constraints that were created (by swapping) to avoid deadlock.
- removeConstraints(tlsID, tripId, foeSignal, foeId): remove constraints with the given values. Any combination of inputs may be set to "" to act as a wildcard filter """
- updateConstraints(vehID, tripId=""): remove any constraints related to the given tripId if the vehicle with the given vehID no longer passes the respective rail signals (i.e. after rerouting).
- if tripId is the empty string, the current tripId param of the vehicle is used
- if the tripId would have changed in the part of the route affected by rerouting, the function must be called once for each individual tripId involved.
The length of railway carriages, locomotive and the gap between the carriages can be configured using the following generic vType parameters:
These parameters control the appearance of trains in sumo-gui when drawing vehicles with the style 'simple shapes'.
By default, railway tracks are shown in sumo-gui in a distinct style:
- tracks and crossties are rendendered instead of filled shapes when zoomed in (configurable with visualization setting show rails)
- turning arrows are not shown
- the white bar that indicates right of way is not drawn
- junction shapes are not shown and the internal edges that make up a railway switch are always visible (can be changed setting junction coloring to any scheme other than uniform)
- traffic light (rail signal) indicators are drawn with an offset to the right to indicate the direction in which the signal is applied
- for each pair of edges that make up a bidirectional track, only of is drawn to avoid artefacts. To make both edges visible (i.e. for selecting or otherwise interacting with that edge), the visualization setting spread bidirectional railways/roads in the Streets tab can be activated
- the pre-defined gui setting scheme "Rail" can be used to activate the following settings that are helpful for diagnosing a railway simualation
- spread bidirectional railways/roads
- coloring junctions by type (making rail signals more visible)
- constant junction size (making rail signals bigger when zoomed out)
- show lane direction (direction is otherwise hard to judge for uni-directional tracks)
- vehicles are shown as simple shapes to ensure articulated train shapes follow curves
Road networks are most often modelled according to their actual layout in cartesian space. Such networks have all distances and angles in the same proportions as those of geographical map.
In the railway domain it is often useful to work with schematic (also called abstract) networks instead of (or in addition to) geographical networks. Such abstract networks can make it easiert so see all tracks and switches on a single screen without zooming and panning. SUMO supports working with abstract maps in the following ways:
- all roads and tracks can have a custom "length" value that differs from their visual length. This allows to separate the visualization of the network from it's simulation behavior (w.r.t. distance travelled).
- sumo-gui supports loading an abstract map of a network along with a geographical map by using options -n geo.net.xml -N abstract.net.xml. The two networks must have the exact same topology and may only differ in their geometry.
- The user may switch between the visualization of either geometry via the hotkey CTRL+K or by setting Street visualization setting secondary shape.
- All outputs that include geometry information (i.e. fcd-output) will be according the the network loaded with option -n
- the tool abstractRail.py can be used to convert geographic rail networks in abstract rail networks
- Error checking for railway schedules can be done with the tool checkStopOrder.py
- Individual rail cars / coupling / uncoupling cannot currently be modeled