ROS-based Challenge Agent

Interfacing CARLA from ROS is normally done via CARLA ROS Bridge. In Challenge Mode this bridging functionality is provided by a RosAgent. It uses the same topics and message-types for the sensors but does not publish tf-transformations.

Requirements

roscore is expected to be running in the docker container. Please adapt your entrypoint.

Setup

To enable your stack within challenge mode, the following steps need to be taken:

Define Sensor Setup
Define Startup

Define Sensor Setup

Derive from RosAgent and implement the sensors() method.

from srunner.autoagents.ros_agent import RosAgent

class MyRosAgent(RosAgent):

    def sensors(self):
        return [ <sensor-definition> ]

As an example for the sensor definition, see HumanAgent.py.

Define Startup

The startup of the stack is done within the shell script $TEAM_CODE_ROOT/start.sh. Therefore the environment variable TEAM_CODE_ROOT must be set.

RosAgent takes care of executing and monitoring. The script shall remain running as long as the stack is active.

Example for start.sh

#!/bin/bash -e
roslaunch $TEAM_CODE_ROOT/challenge.launch

Testing

In general, the challenge execution is headless. For diagnosis you're still able to use ros-tools like rviz or rqt. Additionally you can use carla_manual_control from the carla_ros_bridge for visualization (and also controlling the vehicle).

ros_agent.md 1.6 KB لینک دائمی تاريخچه خام