f1tenth notes

• In order to access data from usb devices, ensure that the device is plugged in and you are accessing the correct port (Use ls /dev/tty* to check, you should look for /dev/tty/USB* or /dev/tty/ACM*)
• You also need to add your user to the dialout group by running sudo adduser example_username dialout
• Need to make python files executable, ex. chmod +x example.py
• Building (using catkin_make) for the first time takes a long time due to building costmap
  • The navigation and teb_local_planner packages are cloned in this repo to allow for editing and debugging, but this does cause builds to take far longer than when using the compiled binaries
• Use the tag output="screen" when calling ROS Nodes from a launchfile to pipe the output to the console
• Modifying the accepted data in the EKF Param file (ex. odom0_config param), can cause the filter to not output certain odometry values
• Adding code below to your CMakeLists.txt file in ~/catkin_ws/src allows IDEs such as clion load all the CMakelists files below the current directory which allows for helpful autocorrect, error-checking, and go-to-definiton features

  FOREACH(subdir ${SUBDIRS})
  ADD_SUBDIRECTORY(${subdir})
  ENDFOREACH()```
  

• Adding include_directories(~/catkin_ws/devel/include) to the same file will allow IDEs to view any custom message files that your packages create
  • Add these lines directly above the catkin_workspace() line at the end
• The robot_localization package provides a simple interface for an EKF, or UKF and in our case, uses data from the rf2o_laser_odometry package and the IMU to estimate the current state (in the odom frame, of type nav_msgs/Odometry)
  • In some cases when certain inputs are set to false for the UKF, the package will output no useful odometry data (All zeroes) without issuing a warning message. Setting all values to true in the odom/imu/twist config can reveal if this is an issue.
• The gmapping package uses the filtered odometry data from the EKF (robot_localization pkg) along with the lidar scan data in order to build a dynamic map (occupancy grid) around the robot
• The navigation stack works by creating two separate costmaps, and two separate planners
  • The Global Costmap takes in a map (Occupancy Grid) from the gmapping package and uses it to dynamically build a costmap over time, designating areas that are free, unknown, or occupied (designated by values between 0-255)
  • The Global Planner takes this data and plans a global path using kinematic constraints and a goal state (x, y, theta)
  • The local costmap creates a smaller costmap that is frequently updated, directly from lidar data (doesn't use gmapping and is always centered on the robot's current position)
  • The local planner (in our case using the teb_local_planner as a plugin to move_base), uses the local costmap along with the global plan to create a short term path and series of future poses along with the neccessary linear and angular velocities (geometry_msgs/Twist) to reach the goal
• SBPL Global Planner
  • Uses a .mprim file generated in by matlab that describes that kinetmatic constraints of the car. Detailed instructions can be found here
    • The .mprim file specifies possible next states and uses a weight to specify the preferred transformation
  • Two packages are required to use SBPL. sbpl_latice_planner allows move_base to interface with the sbpl library. The sbpl package contains the updated library and all of the planner logic.
• TEB Local Planner
  • Info about the robot footprint params can be found here
• Drivers
  • Razor IMU
  • RPLidar A2
  • ZED Camera (Only outputs raw video streams, depth sensing requires Nvidia CUDA and a different driver)
  • Teensy To change message files for the teensy, you need to add a msg file in f1tenth/msg, update add_message_files in CMakeLists, then follow the instructions in the f1tenth teensy readme. You must cmake, and rebuild the rosserial libraries folder.
• Launch files
  • An sh file is used when running code on the robot so that the imu, lidar, and teensy can startup before recording data or running navigation
• Paramaters
  • .yaml files can be used to define ros parameters.
    • Indents restrict the scope of the parameters, only allowing access to node handles whose name matches the name at the top of the indented block
• Rosbag
  • To create a new bag file from an existing one up can use the install.sh script
  • For faster updates, go to update settings and change main server to mirrors.ocf.berkeley.edu
  • The script will disable IPv6 (also speeds up updates), updates and upgrades all packages, installs the full desktop version of ROS Kinetic, creates a catkin workspace in ~/catkin_ws, clones the f1tenth repo, adds shortcuts to the user bash profile, edits the hosts file, runs catkin_make, then installs VS Code and the python & cpp extensions (along with pip and pylint), and opens VS Code
  • Only use the script on Ubuntu Xenial (16.04)
• An example of a working tf tree is shown in the other dir
• Bash profile
  • Below are some useful commands to add to your bash profile
  • Source ROS in every terminal window (line 1 & 2)
  • define the ROS Master location and ROS IP (line 3 & 4) to point to the robot
  • Define the ROS Master as local (line 5 & 6) to not connect to ROS on the robot if desired
  • Define an alias to easily view the tf tree (line 7)
  • Define an alias to install dependencies and runs catkin_make (line 8)
  • Define an alais to run catkin_clean and catkin_make (line 9)
  • Define an alias to easily ssh into the robot with X11 display server enabled (line 10)

    source ~/catkin_ws/devel/setup.bash
    export ROS_MASTER_URI=http://192.168.1.1:11311
    export ROS_IP=192.168.1.110
    #export ROS_HOSTNAME=localhost
    #export ROS_MASTER_URI=http://localhost:11311
    alias tf='cd /var/tmp && rosrun tf view_frames && evince frames.pdf &'
    alias cm='cd ~/catkin_ws && rosdep install --from-paths -i -y src -r && catkin_make && cd -'
    alias cclean='cd ~/catkin_ws && catkin_make clean && catkin_make && cd -'
    alias log='ssh -X cyber-physical@cyber-physical'
    

• Useful Repos
  • https://github.com/mrsd16teamd/loco_car/
  • https://github.com/Hypha-ROS/hypharos_racecar
  • https://github.com/mlab-upenn/f1tenthpublic/tree/master/code
  • https://github.com/tysik/obstacle_detector
  • https://github.com/ceezeh/UnmannedNavigationSystem