The main way to interact with a running simulation is TraCI which gives the complete flexibility of doing cross-platform, cross-language, and networked interaction with sumo acting as a server. To allow coupling with client code, bindings must be provided for each client language. Historically, this has led to a larger number of client implementations with varying levels of API completeness and only the python client was kept up-to-date by the core SUMO development team.

The avoid this problem, Libtraci is provided as a SWIG compatible C++ client library that is fully compatible with Libsumo. It even uses the same header files as Libsumo.

  • C++ interface based on static functions and a few simple wrapper classes for results which can be linked directly to the client code
  • Pre-built language bindings for Java and Python (using SWIG)
  • Future support for other programming languages via SWIG

Unlike Libsumo, Libtraci allows

  • multiple clients
  • running with sumo-gui


The following things currently do not work (or work differently than with the TraCI Python client):

  • subscriptions that require additional arguments (except for vehicle.getLeader)
  • stricter type checking
    • the pure Python TraCI client sometimes accepts any iterable object where Libtraci wants a list
    • pure Python may accept any object where Libtraci needs a boolean value
  • there is no cleanup / waiting for the started subprocess (sumo)

Building it#

If swig and the developer packages for your target language (e.g. Python or Java) are installed the build should be enabled by default. For the Python bindings you will get a and a (or .pyd on Windows) in SUMO_HOME/tools/libtraci. For Java the jar and .so (or .dlls) are placed in the bin dir. Please add the bin or tools dir to your relevant search paths.

Using libtraci#



There is no advantage in using libtraci instead of the standard python traci library. The method below is mostly used for testing libtraci.

import libtraci
libtraci.start(["sumo", "-c", "test.sumocfg"])

Existing traci scripts can mostly be reused by calling

import libtraci as traci

In case you have a lot of scripts you can also set the environment variable LIBTRACI_AS_TRACI to a non empty value which will trigger the import as above.


Example Code (test.cpp)#

#include <iostream>
#include <libsumo/libtraci.h>

using namespace libtraci;

int main(int argc, char* argv[]) {
    Simulation::start({"sumo", "-n", ""});
    for (int i = 0; i < 5; i++) {

compiling on Linux (make sure SUMO_HOME is set and sumo has been built)#

g++ -o test -std=c++11 -I$SUMO_HOME/src test.cpp -L$SUMO_HOME/bin -ltracicpp

running on Linux#



You might want to use the available Maven package.

Example Code (

import org.eclipse.sumo.libtraci.*;

public class APITest {
    public static void main(String[] args) {
        Simulation.start(new StringVector(new String[] {"sumo", "-n", ""}));
        for (int i = 0; i < 5; i++) {

compiling on Linux#

make sure SUMO_HOME is set and the jar / so / dll files are available

javac -cp $SUMO_HOME/bin/libtraci-1.8.0-SNAPSHOT.jar

running on Linux#

java -Djava.library.path=$SUMO_HOME/bin -cp $SUMO_HOME/bin/libtraci-1.8.0-SNAPSHOT.jar:. APITest