CAD2CAV: Computer-Aided Design for Cooperative Autonomous Vehicles

This is the github project for the F1Tenth Independent Study Projects 2021. In this project we want to plan the paths for multi-agent given a prior of an environment like a blueprint of the building’s floor plan.

For Developers: This repos has not set up CI tools. Please make sure clang-format is properly in use when contributing to the code base, in order to keep coding style consistent.

System Requirements

• Linux Ubuntu (tested on 20.04 LTS)
• ROS Noetic
• GCC 5+/Clang 5+ (support for C++17)

Software Requirements

• Boost 1.71
• LibConfig++ 1.5.0
• Eigen 3.3.7
• OpenCV 4.5.1
• Google OR-Tools 9.0
• OSQP 0.6.2
• OSQP-Eigen 0.6.3
• METIS 5.1.0
• ROS Integration for Cartographer SLAM
• Protocol Buffers (Protobuf) 3.17.3
• shineyruan/cad2cav_common

(Deprecated, contact author for update) This project also depends on another UE4 software that provides initial landmark locations for the planner to plan. One should also have the following:

• shineyruan/unreal_levine_4
• Autodesk Revit 2019
• Autodesk AutoCAD 2019
• Unreal Engine 4.23.1

This project also has an on-board part that is only meant to be compiled on an F1TENTH vehicle. To checkout the code, please refer to:

• shineyruan/cad2cav_onboard

First run and install the following packages from Ubuntu repository:

sudo apt install libconfig++-dev libboost-all-dev libmetis-dev google-mock libgmock-dev

Then install other dependencies (OpenCV, Google OR-Tools, OSQP, OSQP-Eigen, Protobuf) by building from source manually.

NOTE. The graph_partitioner package provides 2 different ways of building Google OR-Tools. Please check out this README for more details.

Installation

1. This repo serves as a collection of ROS packages and you can plug it directly into any Catkin workspaces.

mkdir -p ~/catkin_ws/src
cd ~/catkin_ws/src
catkin_init_workspace
git clone --recursive https://github.com/mlab-upenn/ISP2021-cad2cav.git
git clone --recursive https://github.com/shineyruan/cad2cav_common.git
cd ..

2. Please follow the official documentation for Cartographer SLAM to install the latest cartographer_ros package in the src/ folder.

3. To build the project without pre-installing Google OR-Tools for graph partitioner (which would probably take about 5 more minutes), run

cd ~/catkin_ws
catkin_make_isolated --install --use-ninja -DBUILD_ORTOOLS=ON
source install_isolated/setup.bash

otherwise, if the versions of other software strictly matches the requirements above, one can also use the pre-built OR-Tools binaries in graph_partitioner/3rdparty:

cd ~/catkin_ws
catkin_make_isolated --install --use-ninja -DBUILD_ORTOOLS=OFF
source install_isolated/setup.bash

Note that this project contains multiple large repositories (Cartographer SLAM, Google OR-Tools) and it might take a considerable amount of time to build for the first time (~7 minutes on Ryzen 7 3700X, ~17 mins on i7-8550U). It is recommended to use Ninja over Make for faster compiling speed. catkin_make_isolated is also required for Cartographer SLAM as it contains non-ROS packaged subdirectories.

4. (Deprecated) Developers of this project should also have Unreal Engine 4.23.1 installed in the system. From this point onwards, we assume that your UE4 is cloned and installed in directory ${UE4_ROOT}.

cd ~
git clone --recursive https://github.com/shineyruan/unreal_levine_4

One can check the installation of UE4 by trying to open the project in UE4 Editor:

cd ${UE4_ROOT}/Engine/Binaries/Linux
./UE4Editor ~/unreal_levine_4/Levine_4.uproject

5. To run the system on an F1TENTH car, one should make a Catkin workspace on the car and clone the following repo on-board:

mkdir -p ~/catkin_ws/src
cd ~/catkin_ws/src
catkin_init_workspace
git clone https://github.com/shineyruan/cad2cav_onboard.git
cd ..
catkin_make_isolated

Running the code

User Waypoint Registration

Users can click to specify planning waypoints by running the following application:

rosrun map_service waypoint_registration_node

Coverage Sequence Generation

After getting desired waypoint, one can run the following to generate initial routing scheme for multiple vehicles:

roslaunch auto_mapping_ros coverage_sequence_creator.launch

Coverage Sequence for Car 1	Coverage Sequence for Car 2

Run Planning and Control

Then one can run the planning and control node to navigate F1Tenth race cars by running the following:

(In this repo)

rosrun map_service revit_map_test
roslaunch auto_mapping_ros auto_mapping_ros_real.launch

(On the race car, shineyruan/cad2cav_onboard)

roslaunch particle_filter localize_nomapserver.launch
roslaunch racecar teleop.launch

Press the RB button on the Logitech controller to put race car in autonomous navigation mode.

The screenshot below shows the FMT* planing algorithm open sample set (in green) and the resulting planned path (in red) for the next waypoint. The coverage sequence for this race car is shown in blue.