Skip to content

Latest commit

 

History

History
131 lines (112 loc) · 5.47 KB

notes.md

File metadata and controls

131 lines (112 loc) · 5.47 KB

Notes to use map and localization with Jackal

Communicating with the Jackal

  1. SSH into the jackal

  2. Setup an external machine (If you want to send commands or communicate with jackal remotely. Not necessary for mapping)

    • Create a file named remote-jackal.sh
    • In the file add the following lines
    export ROS_MASTER_URI=http://jackal5:11311  # Jackal's hostname
export ROS_IP=199.168.1.170               # Your laptop's wireless IP
    • Get external machine IP with hostname - I
    • Source the newly created file with source remote-jackal.sh
    • Check if you can echo topics. If not you may have to match the host with the ip
      • To add run sudo nano /etc/hosts
      • Add jackal IP to list of hosts 199.168.1.25 jackal5

Mapping

  1. Launch the velodyne

    • Terminal 1: roslaunch velodyne_pointcloud VLP16_points.launch

  2. To map

    • Terminal 2: roslaunch jackal_navigation gmapping_demo.launch scan_topic:=/scan
    • Drive around slowly and map the area
    • Terminal 3: rosrun map_server map_saver -f labmap
      • labmap is the name of the map file and yaml that will be created in whatever directory you are in
    • To view the map in progress: rosshow /map WARNING: rosshow sometimes kills the wifi connection

  3. To save the map on the external machine run two commands:

    scp [email protected]:/home/administrator/rhodes.pgm rhodes.pgm
scp [email protected]:/home/administrator/labmap.yaml labmap.yaml
[email protected]:/home/administrator/nri_workspace/src/safety_switch.py safety_switch.py
    • To view on the external machine run: roslaunch jackal_viz view_robot.launch config:=gmapping
      • This is working for me but its not really needed
    • View map on matlab (if graphics problem: matlab -softwareopengl) Here you can make a line segment map and choose waypoints.

  4. Tips for mapping:

    • Jackal5 will sometimes turn off after a random time. Maybe 20-30 minutes. Not sure why
    • Be careful of loop closures. Play with the parameters to get something that looks good
    • People passing by seems to be ok. I would try and constantly move and stay further than the clipping range (5m by default)

  5. Places where the map is hardcoded and should change after remapping

    • Localization run command (see below)
    • Specification uses the walls, waypoints in runEvBasedSTL.py
    • Human pose projection in cleaner.py

Localization

  1. Drive to the origin of your map. Reboot.
  2. Launch the velodyne
    • Terminal 1: roslaunch velodyne_pointcloud VLP16_points.launch
  3. To map
    • Terminal 2: roslaunch jackal_navigation amcl_demo.launch map_file:=/home/administrator/rhd3_withoutbox_map6.yaml scan_topic:=/scan
    • For Upson 5th floor, use full_corridor_first_pass. Issue: The origin for this map is inside of the room, but we want to start in the hallway.
    • For Rhodes 3rd floor, use rhodesFinal. David says this one is not great, so we should remap.
    • For Rhodes 3rd floor with added boxes, use rhodes_3rd_boxes. David says this one is not great, so we should remap.
    • Note: It seems possible to start from not the origin, but the default initial pose is (0,0) 0 yaw
  4. To see pose estimate,
    • Terminal 3 (amcl global estimate): rosrun tf tf_echo map base_link
    • Terminal 4 (jackal dead reckoning): rosrun tf tf_echo odom base_link
  5. Topics to record with rosbag record:
/tf
/tf_static
/scan
/odometry/filtered
/jackal_velocity_controller/cmd_vel

Running the MK UWB Tracker

  1. Navigate to the PC Shell folder and run the tracking script
    • cd ~/mk_uwb/Kit\ SR150_040\ Ext\ pack\ -\ USB\ v3.6/Software/MK\ UWB\ PC\ Shell/MK\ UWB\ PC\ Shell\ v1.1.0/
    • python3 track_next_device_ros.py /dev/ttyUSB0
  2. Relevant lines of code if you want to modify things
    • Topic name: pc_shell/device_tracker_ros.py, end of the constructor. Topic name is currently /uwb/relative_tag_location
    • Pose computation: pc_shell/device_tracker_ros.py, in method rng_aoa_to_tf_stamp. Currently puts the distance as x coordinate, 0 as y coordinate Code is in this private repo=. Contact [email protected] for access.

Running the Human Tracking Visualize: rviz -d ~/Documents/nir_project/humandetect.rviz

Running the Zed camera ROS node

  1. SSh in and, in separate terminals, run

    roslaunch zed_wrapper zed.launch
    rosshow /zed/zed_node/left/image_rect_color

  2. For an example ROS node that listens to an image and outputs a string, in separate terminals run

    python3 ~/touch_sense/image_republisher.py
    rostopic echo /zed_img_shape

Running the Specification

  1. Find the asl-laptop-1 and login
    • Username: David
    • Pass: ASLAdmin!
  2. Run the localization on the robot (see above)
  3. The code for running specifications is in the this repo: https://github.com/davidgundana/Event-based-STL.git
  4. In a new terminal on the local machine, setup the environment and connect to Jackal5 
    export ROS_MASTER_URI=http://192.168.1.44:11311
conda activate nri
source ~/Documents/nir_project/masterOnJackel.bash
  5. In the same terminal run the specification 
    cd ~/catkin_ws/src/Event-based-STL/FinalPackage
python3 runEvBasedSTL.py