Grand Prix Metaverse Autonomous Driving Challenge

Contest website: https://contests.cis.um.edu.mo/meta_racing_2023

Prerequisite

Platform:
- Python >= 3.8
- Ubnutu 20.04
- ROS Noetic: http://wiki.ros.org/noetic/Installation/Ubuntu
- Download the CARLA-MAP: Google Drive / Baidu Pan (提取码: absd)

sudo apt-get install ros-noetic-carla-msgs ros-noetic-ackermann-msgs ros-noetic-derived-object-msgs

pip install numpy cvxpy transforms3d pygame

Environment Variables

export CARLA_ROOT=<path-to-carla>
export PYTHONPATH=$PYTHONPATH:$CARLA_ROOT/PythonAPI/carla/dist/carla-<carla_version_and_arch>.egg:$CARLA_ROOT/PythonAPI/carla

For example:

export CARLA_ROOT=~/grandprix/map
export PYTHONPATH=$PYTHONPATH:$CARLA_ROOT/PythonAPI/carla/dist/carla-0.9.13-py3.8-linux-x86_64.egg:$CARLA_ROOT/PythonAPI/carla

Start CARLA server

cd $CARLA_ROOT
./CarlaUE4.sh

Build the ros-bridge

cd bridge
catkin_make
source devel/setup.bash

Note: you may need to run

cd src/carla_ros_bridge/src/carla_ros_bridge
chmod +x bridge.py

Spawn agents and Start CARLA-ROS-Bridge

roslaunch carla_ros_bridge test.launch

Complete the file motion_planner.py in submit

core.py: The core file to implement the MPC controller
ref_path.txt: The reference path file
run.py: The file to run the ROS node
answers/example_motion_planner.py: The example answer file that you can refer to
answers/motion_planner.py: The file should to be completed

Run example answers

First, put the python file in the folder submit/answers. Then, run the following command:

python submit/run.py -f YOUR_FILE_NAME

For example, to run the MPC path tracking, the command should be:

python submit/run.py -f 'example_motion_planner_mpc.py'

For example, to run the pure pursuit path tracking, the command should be:

python submit/run.py -f 'example_motion_planner_purepursuit.py'

Testing with NPC agent

In this challenge, you have to deal with situations where other vehicles appear on the track at the same time. The ego-vehicle should avoid or change lanes independently to complete the race. Therefore, we have a script to generate a NPC vehicle in a random spawn point.

python testing/vehicle_random_generate.py

change the control mode of the NPC vehicle

# ego control mode, "set_transform" OR "ackermann_control" OR "control" OR "autopilot" "stop"

CONTROL_MODE = "autopilot"

hanruihua/grandprix