carla_scenario

Python 2.x and 3.x scripts which support you to create scenarios in CALRA, written in python, for human beings all over the world. These scripts enable you to use CARLA Python API via XML file.

Demo

./CarlaUE4.sh
python3 scenario_xml.py -s scenario/town5.xml
python3 manual_control.py --rolename ego_vehicle

Features

Control actors and trafficlight when the ego_vehicle steps on triggers.
Script read XML file and run CARLA Python API (scenario_xml.py) Supports almost all of functions. (spawn, move, kill, pose, traffic light). You can customise it by CARLA Python API.
Visualize the created scenario (check_scenario.py) You can check scenario by projecting spawn points, actor trajectory, kill points.
Supports ROS msg derived_object_msgs/Object.msg (carla_ros_bridge.py). The carla-ros-bridge doesn't support obstacles which added to the CARLA world after launching the carla-ros-bridge. So you need to spawn all of the actors at first. It's not good for long and sophisticated scenarios because it consumes the processing power a lot. These scripts support spawn and kill actors and update them to the ROS message. It saves processing power, good for the long scenarios. And this script can repeat the scenario after the scenario finished.

Requirement

CARLA (0.9.6-0.9.8 are tested)
ROS (option)

Install

pip install -r requirements.txt

Prease confirm that you sourced .egg file provided by CALRA.

Usage

Run Scenario

Run simulator and change map.
spawn ego vehicle with rolename=ego_vehicle

python manual_control.py --rolename ego_vehicle

Run scenario

python scenario.xml -s /path/to/scenario.xml

Run Scenario with ROS

Run simulator and change map.
spawn ego vehicle with rolename=ego_vehicle

python manual_control.py --rolename ego_vehicle

roscore
run scenario with ros

python carla_ros_bridge.py -s /path/to/scenario.xml

The object msg will be published

Create Scenario

Demo scenario is here

0. Get spawn positions

# simulator
./CarlaUE4.sh
# Control HERO and copy it's position to the clipboard for scenario configulation
python3 manual_control.py --filter vehicle*

--filter: walker.pedestrian* or vehicle*

How to copy HERO position to the clipboard

L Button: get actor's location to clipboard for xml file
T Button: get actor's tralsform to clipboard for xml file

1. Set trigger

The scenario is read from top to bottom. If the vehicle misses one trigger, the scenario will stop at the point. Please be careful when deciding the "thres" value The trigger 0 is considered as initial setting. The scenario of trigger 0 is launched when the scenario starts whether the vehicle has stepped on the trigger.

<trigger id="enter_road1" thres="2.0">
  <location>138.851516724,-2.28138899803,0.00914668943733</location>

id: Set as you like, understandable.
thres: The vehicle is considered to have stepped on the trigger when the vehicle is within the range of the value(m) from the trigger position.
location: trigger location on the map

2. Spawn Actors

walker

        <spawn id="ai_walker_road1_0">
            <type>ai_walker</type>
            <probability>20</probability>
            <blueprint>random</blueprint>
            <transform>122.392601013,-8.89567947388,1.11982262135,0.0,-179.000015259,0.0</transform>
        </spawn>

id: Set as you like, understandable.
type: ai_walker is controlled by simulator, walker is controlled by this script (track the specified waypoints).
probability: value for derived_object_msgs::Object::classification_certainty
blueprint: select from HERE or set random.
transform: spawn position

vehicles

        <spawn id="vehicle_road2_0">
            <type>vehicle</type>
            <probability>80</probability>
            <blueprint>random</blueprint>
            <transform>-55.2978057861,26.842716217,0.0971065685153,0.0965993627906,90.5140838623,6.39263889752e-05</transform>
        </spawn>

id: Set as you like, understandable.
type: ai_vehicle is controlled by simulator, vehicle is controlled by this script (track the specified waypoints using pure-pursuit).
probability: value for derived_object_msgs::Object::classification_certainty
blueprint: select from HERE or set random.
transform: spawn position

static objects

        <spawn id="static_road5_0">
            <type>static</type>
            <probability>80</probability>
            <collision_range>10.0</collision_range>
            <invincible>true</invincible>
            <blueprint>static.prop.plasticbag</blueprint>
            <transform>-12.8004541397,94.2045593262,0.951499938965,0.0,78.3999176025,0.0</transform>
        </spawn>

id: Set as you like, understandable.
type: static
probability: value for derived_object_msgs::Object::classification_certainty
collision_range: The object disappears when the vehicle is within the value(m) if invinsible is set to true.
invinsible: If false, the object disappears. If true, the object becomes invinsible.
blueprint: select from HERE or set random.
transform: spawn position

3. Control Actors

Control ai actor

        <move id="ai_walker_road1_0">
            <waypoint speed="3.0"/>
        </move>

waypoint: specify only speed (km/h)
Control actor

       <move id="vehicle_road2_0">
            <waypoint speed="20.0">-57.9040222168,49.9933433533,1.1593503952</waypoint>
            <waypoint speed="30.0">-40.0210914612,50.7340812683,1.1593503952</waypoint>
        </move>

waypoint: the goal point of actors (transform of point)

4. Kill Actor

        <kill id="ai_walker_road1_0"/>
        <kill id="vehicle_road2_0"/>

5. Control trafficlight

When the vehicle steps the trigger, the light turns to the color

        <trafficlight>
            <location>-38.7292823792,-8.71572208405,1.11481559277</location>
            <state>red</state>
            <time>30.0</time>
        </trafficlight>
        <trafficlight>
            <location>-60.142124176,-9.56110668182,1.12300264835</location>
            <state>green</state>
            <time>30.0</time>
        </trafficlight>
        <trafficlight>
            <location>-59.4050865173,11.8945960999,1.14123988152</location>
            <state>red</state>
            <time>30.0</time>
        </trafficlight>
        <trafficlight>
            <location>-39.0418167114,11.9720258713,1.15340161324</location>
            <state>green</state>
            <time>30.0</time>
        </trafficlight>

location: the traffic light pole location (the script finds pole within 3m from specified position) state: red, green, yellow time: Time to keep the color

Check Scenario

Run simulator with server window (to get bird eye view), then change map python carla-0.9.8/Python/util/config.py -m Town05
Run check scenario

python check_scenario.py -s /path/to/scenario.xml

You can check the scenario. (Consistency check function will be supported...)

MarinBao/carla-scenario