robot_pkg

Test Vehicle

All software parts are running in a docker container. Either build the image from source or download it from docker hub.

Note, the image does only work on ARM- Architecture!

Building the docker image:

1. Clone this repository
2. On the Jetson Nano go to Ackerbot/ackerbot_real_ws/src/robot_pkg
3. Open Terminal here
4. Close all other Applications (to have enough system memory, when building on Nano)
5. sudo docker build -t dschori/ackerbot:real -f Dockerfile .

Starting the docker container:

Only the first time:

To copy the build files and directories to the local file system.

1. Build the workspace inside the container with: catkin build -j2
2. Then, copy the build files inside the container to the local file system with:

sudo docker run -d --name ackerbot dschori/ackerbot:real

sudo docker cp ackerbot:/workspace/catkin_ws/build ~/Ackerbot/ackerbot_real_ws/build
sudo docker cp ackerbot:/workspace/catkin_ws/devel ~/Ackerbot/ackerbot_real_ws/devel
sudo docker cp ackerbot:/workspace/catkin_ws/logs ~/Ackerbot/ackerbot_real_ws/logs  

sudo docker rm -f ackerbot  

Then:

sudo docker run -it --rm --privileged --net=host --name ackerbot \
-v ~/Ackerbot/ackerbot_real_ws/src/robot_pkg:/workspace/catkin_ws/src/robot_pkg \
-v ~/Ackerbot/ackerbot_real_ws/build:/workspace/catkin_ws/build \
-v ~/Ackerbot/ackerbot_real_ws/devel:/workspace/catkin_ws/devel \
-v ~/Ackerbot/ackerbot_real_ws/logs:/workspace/catkin_ws/logs \
dschori/ackerbot:real bash

When inside the docker container source the build with: source devel/setup.bash

SLAM:

1. Starting the sensors launch file:

This starts the Velodyne VLP-16 and the Realsense T265 Sensor

Open a new Terminal and go into the running container:

sudo docker exec -it ackerbot bash

Inside the container type:

source devel/setup.bash
roslaunch robot_pkg sensors.launch

2. Starting the mapping launch file:

For the TEB navigation as well as the localization of the vehicle Google Cartographer was used as SLAM algorithm.

To improve the mapping and localization process, IMU and Odometry Data from the Realsense T265 is fed into Cartographer.

This File starts the Google Cartographer SLAM process

roslaunch robot_pkg mapping.launch

3. Driving around:

This File starts the drive controller, one can now drive around. See: Teleop Twist

roslaunch robot_pkg drive.launch

4. Open RVIZ

Open RVIZ with the config file and monitor the nodes you want.

TEB Navigation:

1. Start the Sensors as mentioned in SLAM

2. Starting the localization launch file:

This File localizes the robot in a given map based on the laser scans and odometry data of the T265.

In a new Terminal run:

roslaunch robot_pkg localization.launch

3. Start the TEB Navigation:

This File launches the TEB Navigation packages

In a new Terminal run:

roslaunch robot_pkg teb_navigation.launch

4. Start Rviz

Start RVIZ with the config file and set a "2D Nav Goal"

Reinforcement learning based Navigation:

1. Start the Sensors as mentioned in SLAM

2. Starting the localization launch file:

This File localizes the robot in a given map based on the laser scans and odometry data of the T265.

In a new Terminal run:

roslaunch robot_pkg localization.launch

3. Start RL Navigation

Open the Inference Notebook to load the trained Agent, set a target and start the navigation.