Skip to content
Go to file


alt text

Software for the AutoRally research platform.

AutoRally Platform Website

AutoRally Youtube Channel

Platform Paper


  title={AutoRally: An Open Platform for Aggressive Autonomous Driving},
  author={Goldfain, Brian and Drews, Paul and You, Changxi and Barulic, Matthew and Velev, Orlin and Tsiotras, Panagiotis and Rehg, James M},
  journal={IEEE Control Systems Magazine},


We welcome bug fixes, ehancements, new features, and feedback!

Please submit pull requests to the melodic-devel branch that conform to the ROS C++ Style Guide. We use Gitflow, so master branch is reserved for releases.

Setup Instructions


  1. Install Prerequisites
  2. Clone repository
  3. Install AutoRally ROS Dependencies
  4. Compilation/Running
  5. Generate Documentation
  6. Test Setup in Simulation
  7. Autonomous Driving in Simulation

1. Install Prerequisites

  1. Install Ubuntu 18.04 64-bit

  2. Install required packages

    sudo apt install git doxygen openssh-server libusb-dev texinfo

    ROS Melodic only supports Python 2.7. Before installing Python packages, you need to ensure that python points to Python 2.7, e.g., by setting up a Python 2.7 conda environment:

    conda create -n my_ros_env python=2.7
    source activate my_ros_env
    conda install defusedxml
    conda install -c jdh88 rospkg

    The following tools are recommended, but not required for this project.

    • cutecom
    • cmake-curses-gui
    • synaptic
    • python-termcolor

    sudo apt install cutecom cmake-curses-gui synaptic python-termcolor

  3. Install ros-melodic-desktop-full

  4. Install MPPI Dependencies (if you have a GPU and will run MPPI)

    The core idea behind MPPI is to sample thousands of trajectories really fast. This is accomplished by implementing the sampling step on a GPU, for which you will need CUDA. We also use an external library to load python's numpy zip archives (.npz files) into C++.

  5. Install gtsam

    Follow the gtsam Quick Start guide to clone and install the develop branch of gtsam.

    Instead of cmake .., use:


    Once install is complete, make sure linux can see the shared library:

    sudo ldconfig

  6. Update to latest version of gazebo 9.XX

    You will want to be on the latest version of gazebo 9.

2. Clone or Fork Repositories

Get the autorally repository in a catkin workspace. The suggested location is ~/catkin_ws/src/, but any valid catkin worskspace source folder will work. We suggest forking over cloning if you will be working with the code.

Also clone the IMU code and Pointgrey camera drivers into the same catkin workspace:

git clone
git clone

3. Build Pointgrey Camera Driver

Since there are no pre-built drivers for Melodic, follow these instructions to build the driver yourself.

  • The SDK can be downloaded here. The file to download is flycapture2-
  • You may need to run sudo apt --fix-broken install after installing the suggested packages and before running sudo sh

4. Install AutoRally ROS Dependencies

Within the catkin workspace folder, run this command to install the packages this project depends on.

rosdep install --from-path src --ignore-src -y

5. Compilation & Running

First, check your Eigen version with pkg-config --modversion eigen3. If you don't have at least version 3.3.5, upgrade Eigen by following "Method 2" within the included INSTALL file.

Then, to compile and install, run catkin_make from the catkin workspace folder. If your version of CUDA does not support gcc-7, you may need to use


Due to the additional requirement of ROS's distributed launch system, you must run

source src/autorally/autorally_util/

before using any AutoRally components. See the wiki for more information about how to set this system up for distributed launches on your vehicle platform.

Note: If you are unfamiliar with catkin, please know that you must run source catkin_ws/devel/ before ROS will be able to locate the autorally packages (and thus you must run this before sourcing This line can be added to your ~/.bashrc file so that it is automatically run on opening a terminal.

6. Generate Documentation

You can generate or update code documentation by running doxygen in autorally/.

To view code documentation open autorally/doc/html/index.html in a web browser.

7. Start the AutoRally Simulation to Test Configuration

roslaunch autorally_gazebo autoRallyTrackGazeboSim.launch

You can use a USB gamepad to drive the simulated platform around. On startup, the runstop message published by the joystick node is false. Press any of the buttons by the right stick (normally labelled X, Y, A, B or square, triangle, X, circle) to toggle the published value.

Verify runstop motion is enabled by looking at the runstopMotionEnabled field in the /chassisState topic (rostopic echo /chassisState).

If you aren't using a gamepad, you will have to configure another source of runstop information for the platform to move:

  • Comment out the line <include file="$(find autorally_control)/launch/joystickController.launch" /> near the end of autorally_gazebo/launch/autoRallyTrackGazeboSim.launch

  • rosrun rqt_publisher rqt_publisher

and configure rqt_publisher to publish a message to topic /runstop of type autorally_msgs/runstop at 1 Hz with sender set to rqt_publisher and motionEnabled set to true.

  • Verify that runstopMotionEnabled is true in /chassisState topic.

8. Autonomous Driving in Simulation

At the end of this section the robot will be driving autonomously in simulation using controllers available in autorally_control.

Position the robot in the same spot as when the simulation starts and make sure runstop motion should is enabled (set to true).

Start waypoint follower:

roslaunch autorally_control waypointFollower.launch

Start constant speed controller and tell it what speed to drive:

roslaunch autorally_control constantSpeedController.launch
rosrun rqt_publisher rqt_publisher

Configure a publisher on topic constantSpeedController/speedCommand of type std_msgs/Float64 at rate 10 with value of 3 (you can adjust he value once everything is running). The value is the target velocity in m/s, and as soon as you do this the platform should move if motion is enabled.

If the robot turns and hits the barrier it's probably because the state estimator has not converged, so its orientation estimate is incorrect. Just select the track barriers and move them up to allow the robot to continue driving, and the estimator should converge and the vehicle will return to within the barriers.

What's Next

More detailed explanations of the controllers and state estimator can be found on the wiki:

Controlling the AutoRally platform is a tutorial for how your own controller can control the AutoRally platform (in simulation or on hardware).

Running Vehicles in Simulation is a tutorial on how run gazebo in the two different worlds and with multiple vehicles.

If you are configuring a physical AutoRally platform, the next step is to configure the compute box, all of the peripherals, and the launch system. Those instructions are found in the Platform Configuration Instructions.

You can’t perform that action at this time.