BFMC Sim in ROS, autonompus:go Team. Implementation of a smart lane following, important feature detection using yolov5 and intelligent decision making. All the implemented nodes were tested and conected to an autonomous car model.

• Install Ubuntu 20.04.2.0 LTS (Focal Fossa) - Desktop Image. For best performances, avoid using a virtual machine.
• Install ROS Noetic Ninjemys - Desktop-Full Install.

Clone the repository inside the Documents folder (or perform a commit on the repository)

• Open a terminal in ~/Documents/Simulator directory and run the following command:

catkin_make --pkg utils
catkin_make

• Two folders – devel and build – will be generated inside the repository directory.

• Gazebo needs to know where the workspace's packages and models are, so you will need to add them to the gazebo environment variable list. The easiest and cleanest way is to add them to the source file of the simulation, so it won't interfear with other projects:

gedit devel/setup.bash

and add those two lines at the begining of the file (don't forget to change {YOUR_USER} with your actual user name).

export GAZEBO_MODEL_PATH="/home/{YOUR_USER}/Documents/Simulator/src/models_pkg:$GAZEBO_MODEL_PATH"
export ROS_PACKAGE_PATH="/home/{YOUR_USER}/Documents/Simulator/src:$ROS_PACKAGE_PATH"

For each terminal where you plan to use the ROS, you need to source the repository itself:

source devel/setup.bash

• Finally, run the following command in order to start the entire simulation (the simulaiton is heavy, it may take many resources):

roslaunch sim_pkg map_with_all_objects.launch

The simulation can interact directly with the car or it can be used as a stand-alone environment on your PC.

Both the car and the PC running the simulation have to be on the same network and the ssh communication has to be enabled on the RPi. In order for the nodes to be able to communicate between them, some environmental variables have to be set. A script was created for this purpose: On the RPI, you have to run the script as follows:

source src/utils/network.sh PC_IP

On the PC running the simulation, you have to run the script as follows:

source src/utils/network.sh

This setup will make the nodes on your car interact with the roscore from the PC, and you will be able to see the entire setup of the network. The file have to be soruced for each terminal. The core can be swapped to run on the phisical car as well.

Run the simulation on the PC:

roslaunch sim_pkg map_with_car.launch

Run the control nodes of the car on the RPI:

roslaunch utils start_car_virtual.launch

Now the simulator will publish some info on the topics and you can subscribe to them from your car (automobile/image_raw, automobile/localisation, automobile/IMU, automobile/feedback, automobile/semaphores/*)

Now the simulator will subscribe to some info on the topics and you can publish on them from your car (automobile/command)

To check an example of publishing & subscribing, after the simulaiton was started, you can open two other terminals and source the devel/setup.bash inside. on one terminal, execute

rosrun example camera.py

and on another one you can execute

rosrun example control.py

If the setup for interacting with the car is done, the phisical car will also interact with the two scripts, as well as the virtual one. We suggest, if you plan to develop on the PC, to create another workspace for the project itself.

Inside src/models_pkg/camera/model.sdf, you will find the <width> and <height> tags. Modifying them will allow you to change the camera resolution. You don't need to recompile the workspace after this change.

Inside src/models_pkg/rcCar_assembly/model.sdf, you will find the inclusion of <uri>model://camera</uri>. Inside that same <include> tag, there is also a <pose> tag, which you will need to modify according to the SDF format specification in order to change the camera's position. You don't need to recompile the workspace after this change.

<!-- All constants of double data type, with angles in radians and "." (dot) as decimal separator -->
<pose> X Y Z ROLL PITCH YAW </pose> 

• If you want to start the testing from scratch, instead of restarting gazebo, you can only restart the simulation with the help of the ROS services:

$ rosservice call /gazebo/reset_simulatio

• Deactivate GUI from the launch in order to save some computational power;

• Play with rqt tool in order to check the images and other topic messages, frequency, etc.

• You can change the resolution of the map from src/models_pkg/track/materials/scripts/bfmc_track.material. The available maps are under the textures directory. Bigger the resolution, bigger the requirements of the simulation.

• Inside the src/sim_pkg/launch/sublaunchers directory, there is a launch file for each type of elements on the map or groups of elements. You can also spawn some of them separately while the simulator is running, based on the requirements at the moment.

You can also configure your own launching files, you can create a launch file that includes multiple launch files. See the map_with_all_objects.launch file inside the src/sim_pkg/launch as an example.

In order to use the launch files, you need to source the devel/setup.bash