Features that cause rerouting#

There are multiple simulation features that allow routing at simulation time. They are described in the following:

Routing triggered by the vehicle#

This type of routing works by assigning a rerouting device to some or all vehicles. Details are given at Demand/Automatic_Routing.

Incomplete Trips and Flows#

This is a special case of the above method. Vehicles with incomplete routes automatically receive a rerouting device and are rerouted once when entering the network. In some scenarios this is a practical one-shot-approach to route assignment that avoids time-consuming iterative assignment.

Alternative Route Signage#

This is a location based method for triggering rerouting and is described at Simulation/Rerouter.


Using the methods traci.vehicle.changeTarget or traci.vehicle.rerouteTraveltime rerouting is triggered for the specified vehicle.

Alternatively, routes can be computed using traci.simulation.findRoute and applied using traci.vehicle.setRoute.

For persons, the function traci.simulation.findIntermodalRoute can be used to compute simple walks as well as itineraries for public transport.

Alternative Destinations#

By using <rerouter>-definitions, vehicles can be routed to alternative destinations. A different method is to use traffic assignment zones (TAZ). This allows vehicles to change their destination to the best alternative from a list of potential destinations.

Handling of temporary obstructions#

By using <rerouter>-definitions of type <closingReroute allow="..."/> (or <closingLaneReroute allow="..."/> if all lanes are closed) it is possible change road permissions temporarily (i.e. to model construction sites). By default, the simulation will raise an error when loading a route that passes a closed edge or if routing to the destination fails because of an obstruction. By setting a vehicle routing mode that includes bit-flag traci.constants.ROUTING_MODE_IGNORE_TRANSIENT_PERMISSIONS, this behavior can be changed. Instead vehicles will ignore obstructions until reaching the closed edge and then wait there (potentially teleporting after *--time-to-teleport is reached).

Setting the routing mode (to ignore blockage) can be accomplished in the following ways:

  • traci.vehicle.setRoutingMode(vehID, traci.constants.ROUTING_MODE_IGNORE_TRANSIENT_PERMISSIONS)
  • sumo option --device.rerouting.mode 8
  • defining <param key="device.rerouting.mode value="8"> in the vType or a single <vehicle>, <trip> or <flow>.

Travel-time values for routing#

By default, the route with the least travel time is chosen. The travel time depends on the current routing mode (configurable via traci.vehicle.setRoutingMode) or via the explicit routingMode argument to traci.simulation.findRoute.

Routing Mode traci.constants.ROUTING_MODE_DEFAULT#

The following order of steps is taken to retrieve the travel time for each edge. If a step provides data, this is used, otherwise the next step is attempted:

  1. The vehicle retrieves its individual data storage. This can be set and retrieved using the TraCI vehicle methods change edge travel time information and edge travel time information.
  2. The global edge weights loaded using option --weight-files are retrieved.
  3. The global edge weights (set and retrieved via TraCI) using the TraCI edge methods change edge travel time information and edge travel time information.
  4. The minimum travel time (length/allowedSpeed) is used.


The edge weights for cases 1-3 support modeling time-dependent variations in edge travel time. This means future changes in travel time along a vehicles route can be taking into account when routing.

Routing Mode traci.constants.ROUTING_MODE_AGGREGATED#

The smoothed travel times computed for the rerouting device are used. Note, that these can also be modified via TraCI.

Randomized travel times#

Sumo and duarouter support option --weights.random-factor FLOAT. When this option is set, edge weights (i.e. travel times) for routing are dynamically disturbed by a random factor drawn uniformly from [1,FLOAT). The value of FLOAT thereby sets an upper bound on the difference in travel time between the actual fastest path and the routing result. With a value of 2. The result will have twice the travel time / cost of the optimal path in the worst case (In the unlikely event where all edges of the fastest path happen to get a disturbance factor of 2).

Randomized edge weights can be useful in grid networks where there are many paths have the same (or almost the same) travel cost and the default routing algorithm would send all vehicles on the same path. It can also be used to model imperfection in travel time estimation.


By default, runs with the same inputs will still yield the same outcome. To vary the randomness between runs, options --seed or --random must be used.

Special cases#

Routing by Travel time and Edge Priority#

Sometimes it is useful to guide route search with additional information while still taking travel times into account. For this use case the option --weights.priority-factor FLOAT can be used with sumo and duarouter.

When this option is set, the priority value of each edge is factored into the routing decision so that low-priority edges receive a penalty (they appear to be slower) whereas high-priority edges receive little or no penalty. For the option value PriorityFactor, the penalty is computed thus:

  MinEdgePriority : minimum priority value of all edges
  MaxEdgePriority : maximum priority value of all edges
  EdgePriorityRange = MaxEdgePriority - MinEdgePriority

  relativeInversePrio = 1 - ((edgePriority - MinEdgePriority) / EdgePriorityRange)
  effort =  traveltime * (1 + relativeInversePrio * PriorityFactor)

As a consequence:

  • the highest priority edge will get no penalty
  • the travel time of the lowest priority edge is multiplied with 1+PriorityFactor,
  • edges with in-between priorities will get a scaled penalty

Routing by effort#

By default, the objective of the routing algorithms is to minimize the travel time between origin and destination. The travel time can either be computed from the speed limits and vehicle maximum speed, it can be estimated at runtime from the simulation state or it can be loaded from a data file. The latter option allows defining travel times for the future. An example for the relevance of future travel times would be this:

  • a vehicle departs for a long trip at a time where there is no jamming
  • it is known that parts of the network will be jammed later
  • the route of the vehicle computed at departure time can circumvent the jam because the routing algorithm is aware that by the time those edges are reached they will be jammed.

It may be useful to compute routes which minimize some other criteria (called effort) besides travel time (distance, emissions, price, ...). When these quantities are meant to change over time, the routing algorithm needs two kinds of values for each edge:

  • the effort that shall be minimized
  • the travel time for the edge.

The travel time is needed to compute at which time a certain edge is reached so that efforts which change over time can be used correctly.


When the effort values do not change over time, routing by effort can be achieved by loading weight-files with a modified traveltime attribute (the effort value is written into the traveltime attribute) and the option --weight-attribute can be omitted.

When setting the options --weight-file and --weight-attribute, additional routing efforts are read according to the specified attribute. These are only used when calling the TraCI function reroute by effort. The assumed efforts along a vehicles route are time-based values and the time is computed based on the travel time values described above. The effort can also be set using traci.edge.setEffort.


The default effort value is 0 which causes detour routes to be preferred when not loading sensible effort values.

The applications duarouter and marouter also support the options --weight-file and --weight-attribute but they can only be used with one of the weight attributes "CO", "CO2", "PMx", "HC", "NOx", "fuel", "electricity", "noise". However, they will still work as expected when the user loads custom effort values for these attributes.

Routing by distance#

Finding the shortest route rather than the fastest can be achieved by loading suitable effort-data (see #Routing_by_effort) or by setting the speed for all network edges to the same value.

A simpler solution is to define a vehicle type that travels with the same speed on all edges:

<vType id="routeByDistance" maxSpeed="1"/>

and then using that type to find the fastest route:

stageResult = traci.simulation.findRoute(fromEdge, toEdge, "routeByDistance")
shortestDistance = stage.length


Older versions of SUMO do not supply the value stage.length. In this case shortestDistance = stage.travelTime is also correct due to the speed of 1m/s.

Routing Algorithms#

Applications that perform routing (sumo, sumo-gui, duarouter, marouter) support the option --routing-algorithm for selecting among the following values:


Dijkstras algorithm is the simplest and slowest of routing algorithms. It is well suited to routing in time-dependent networks (i.e. when the travel time on an edge depends on the time of day)


This is the default algorithm

astar (A*)#

The A* routing algorithm uses a metric for bounding travel time to direct the search and is often faster than dijkstra. Here, the metric euclidean distance / maximumVehicleSpeed) is used.


By using astar together with the option --astar.landmark-distances <FILE> the ALT-Algorithm is activated. The name ALT stands for: A*, Landmarks, triangle inequality. It uses a precomputed distance table to selected network edges (so-called landmarks) to speed up the search, often by a significant factor.

  • A lookup table can be generated by creating a file with one landmark edge id per line (e.g. landmarks.txt) and then setting the options --astar.landmark-distances landmarks.txt lookuptable.txt.
  • As a rule of thumb using 8-16 edges distributed around the main roads that border the network achieve good ALT-performance. Using more edges is not recommended because each landmark adds a fixed overhead).
  • by using astar together with the option --astar.all-distances <FILE> the A* algorithm is used together with a complete (and often huge) distance table to allow for blazing fast search. This is only recommended for medium sized networks with a high number of routing queries

CH (Contraction Hierarchies)#

Contraction Hierarchies is preprocessing-based routing algorithm. This is very efficient when a large number of queries is expected. The algorithm does not consider time-dependent weights. Instead, new preprocessing can be performed for time-slices of fixed size by setting the option --weight-period <TIME>.

  • When used with duarouter, edge permissions are ignored so this should only be used in unimodal networks.
  • When used with sumo, the computed routes are only valid for the default 'passenger' class.


This works like CH but performs separate preprocessing for every vehicle class that is encountered, thereby enabling routing in multi modal scenarios

Further Options that affect routing#

  • --weights.minor-penalty FLOAT : set a fixed time penalty for passing junctions without having right-of-way (default 1.5s)
  • --persontrip.walk-opposite-factor FLOAT : set penalty speed factor for walking against the flow of vehicular traffic (see pedestrian routing) (default 1)