ELSA - Explore, Localize, Map Simultaeneously and Act

<p align="center"> <img src="images/elsabot_2.jpeg" alt="elsa" width="600"> </p> <p align="center"> <em>ELSA Robot</em> </p>

Project Dependencies

• To install teleop-twist-keyboard

sudo apt-get install ros-melodic-teleop-twist-keyboard

• To install gmapping clone and build this
• To install robot_localization, clone and build this

Project information:

• The project has two packages elsa and ros_arduino_bridge
• Follow the below instructions to build the project

cd ~/catkin_ws`
source devel/setup.bash
catkin_make --only-pkg-with-deps elsa
catkin_make --only-pkg-with-deps ros_arduino_bridge

• Follow the below commands to execute the project

roslaunch elsa elsabot.launch

• Use the below command to do keyboard teleop

rosrun teleop_twist_keyboard teleop_twist_keyboard.py

• To launch gmapping

roslaunch gmapping slam_mapping_pr2.lanch

• To save the map

rosrun map_server map_saver -f /tmp/my_map

• To use robot_localization package for filtered odometry data
  • roslaunch robot_localization ekf_template.launch --> This will be merged to the elsa project, as we are required to edit the EKF parameters according to our requirement

Docs:

• Coordinate Transformations and Robot Dimensions can be found at docs/Robot dimensions and coordinate transformations.pdf
• More info about the references used is provided in the reference section

Finished tasks:

• Publish imu to /imu/raw_data topic
• Publish encoder ticks to /ticks_pub/lwheel_ticks and /ticks_pub/lwheel_ticks topics
• Publish Odometry data to /odom topic
• Configured the ROS_Arduino_Bridge
• Debug the ROS_Arduino_Bridge
  • Solder Arduino Nano
  • Connect the Arduino Nano to the Motor_Driver
  • Edit the scripts for the second USB port (/dev/ttyUSB*)
• Publish encoder ticks to /arduino/lwheel_ticks and /arduino/lwheel_ticks topics
• Publish velocoties to /cmd_vel topic
  • Publish sample Twist msg -- Seems like there's an error/lag
  • Debug the motor_driving
• Finish the base_controller to subscribe to /cmd_vel topic to move the robot_base
• Write a node to Publish raw images to a ROS topic
  • Resolve cv_bridge and opencv4 issue for ROS - the issue is specific to ROS-melodic and OpenCV4 combination
• Transport Pi camaera v2 video stream using sensor_msgs/Image to a ROS topic
• Setup either YDLidar X4 for 2D map creation - needed extra cable to power the LIDAR, since the RPi cannot provide enough power to drive the LIDAR motor
• visualized the tf and odometry in rviz
  • Fix the odometry axis and tf axis alignment - made sure all the axis are aligning in same direction
• rename the coordinate frames of the ydlidar to match with our convention
• fix encoder direction
• fix the motor direction
• Use teleop_twist_keyboard to move the robot -- works perfectly
• Use robot_localization package to filter the Odometry msg using ekf_lozalisation node
  • check if everything is working as expected -- need to tweak covariance matrix for better estimation
• Fix the direction issue with rviz visualization -- Problem is with Odometry calculation -- robot_localization is solving this
• Use slam techniques for mapping - Works
• Tune the PID controller

To Do:

• Use camera data to obtain the images to do SFM and build the 3D_point_cloud_map

Results:

Elsa in Action!

<p align="center"> <img src="images/elsa2.gif" alt="elsa" width="1080"> </p> <p align="center"> <em>ELSA Result</em> </p>

• The below image shows the /base_link and /odom coordinate frames and trasformations between them in rviz

<p align="center"> <img src="images/tf.png" alt="tf" width="1080"> </p> <p align="center"> <em>Coordinate Transformations (TF) in rviz</em> </p>

• The below image shows the robot movement in rviz based on the odometry published to /odom topic

<p align="center"> <img src="images/odom.png" alt="odom" width="1080"> </p> <p align="center"> <em>Odometry of the ELSA Robot in rviz</em> </p>

• The below image shows the laserscan from YDLIDAR X2 being pubilshed to /scan

<p align="center"> <img src="images/laser_scan.png" alt="ls" width="1080"> </p> <p align="center"> <em>Laser Scan in rviz</em> </p>

• The below image shows the TF tree based on coordinate transforms being published to /tf

<p align="center"> <img src="images/frames.png" alt="tf_tree" width="720"> </p> <p align="center"> <em>TF Tree</em> </p>

References:

• The TF tree can be found at elsa/docs/frames.pdf
• The project uses the ROSArduinoBridge Package. However, many modifications are made according to the requirements
• The Project also uses the SparkFun_RedBot_Arduino_Library Arduino Library. This library has been modified a little bit according to our requirement. The updated library can be found in /elsa/elsa/arduino/
• The Projects also uses few insights from the package diff_drive
• The project refers several tutorials from ROS wiki
• The topics are published according to the coordinates frames as per REP103 standards and REP105 standards
• The YDLIDAR uses the package provided by EAI YDLIDAR github
• The Project uses robobt_localization package to fuse imu and encoder-based Odometry data, to provide filtered pose. Technical documentation for the robot_localization package can be found here