Creating a Scenario

Start with a working scenario close to your experiment. A CARLA-only scenario needs an OpenCDA YAML file. Add SUMO assets only for co-simulation, and Artery configuration only when the experiment uses CAPI.

Create the OpenCDA YAML

In the OpenCDA repository:

cp opencda/scenario_testing/config_yaml/v2xp_datadump_town06_carla.yaml \
  opencda/scenario_testing/config_yaml/my_scenario.yaml

Edit the map, agent spawn positions, destinations, sensors and behaviors. The runner merges the file over config_yaml/default.yaml. Vehicle and RSU IDs must be unique within their corresponding agent type. AdvCP uses runtime IDs such as cav-5 and rsu-1. Behavior-service attacks are selected independently through the top-level attacks list.

With CARLA running and CARLA_HOST set in the OpenCDA container:

python opencda.py -t my_scenario --carla-host "$CARLA_HOST" --ticks 200

A scenario in a subdirectory is selected with its relative path, for example -t advcp-scenarios/advcp_check. Validate a CARLA-only run before adding co-simulation. Use Maps and SUMO Assets to inspect coordinates and prepare map assets.

Add SUMO

Create sumo/assets/my_scenario/ in the CAVISE workspace. Reuse a road network from sumo/assets/maps/ that matches the CARLA map, or generate one following Maps and SUMO Assets.

File	Purpose
my_scenario.sumocfg	References routes, network, optional polygons and simulation settings.
my_scenario.rou.xml	Vehicle routes or flows using valid edge IDs from the network.
.net.xml	SUMO road network. Can be shared by scenarios on the same map.
.poly.xml	Optional static geometry for visualization and propagation models.

Paths within .sumocfg are relative to that configuration file. For example, v2xp_datadump_town06_carla.sumocfg references ../maps/Town06/Town06.net.xml rather than copying the network into its scenario directory.

Set the SUMO connection in the OpenCDA YAML (these values can also be inherited from default.yaml):

sumo:
  host: sumo
  port: 3000
  gui: false
  client_order: 2
  step_length: ${world.fixed_delta_seconds}

Start the server with the required client count and then launch OpenCDA with --cosim. Follow SUMO and Artery for the commands. Keep the server’s step length consistent with world.fixed_delta_seconds and ensure routes reference existing edges. Provide traffic or a valid long-lived vehicle so the intended experiment does not end before its clients finish.

Add Artery

For a first CAPI run, use the shared artery/scenarios/capi configuration from SUMO and Artery. It connects to the separate SUMO container at port 3000. This avoids duplicating the network and route files in Artery.

For a custom Artery configuration, use the base capi scenario as a reference. Preserve *.withCAPI = true, the CAPI endpoint, and the TraCI connection settings. With two clients, use client order 1 for Artery and 2 for OpenCDA and start SUMO with --num-clients 2.

A scenario using PosixLauncher starts its own SUMO process. One using ConnectLauncher connects to a server you started separately. Do not mix the two launch methods for the same experiment. If adding a CMake run target, follow artery/scenarios/CMakeLists.txt and rebuild Artery.

Services and Attacks

• Behavior Services describes reusable CAV/RSU services and their configuration.

• Attack Framework describes attacks against those services.

• AdvCP describes attacks against cooperative perception.

Check a baseline without attacks first, then enable one change at a time and compare Recording and Results.