Skip to content

Home

Jump to bottom
Yuxiang Gao edited this page Jul 13, 2022 · 10 revisions

Everything stretch related

Informations

user: hello-robot
passwd: hello2020

The official doc: https://docs.hello-robot.com/ https://docs.hello-robot.com/quick_start_guide/

The specific instructions for the battery charger to avoid damage to the battery: https://docs.hello-robot.com/robot_safety_guide/

The ros related environment variables are managed by direnv and will be automatically loaded when cd into catkin_ws. Without special instruction, all commands on both the local machine and the stretch are run from catkin_ws dir.

Web-based remote control

When the run.launch is launched, you can connect to the robot and control it using the web app: http://intuitivecomputing.jhu.edu/stretch-web-app/#/controller

stretch_ros

The official code are in stretch_ros repo. Currently on the stretch, the remote origin is set to https://github.com/yuxiang-gao/stretch_ros (I would suggest forking this repo and replacing the origin in the local repo with the forked one if you want to make any more updates, otherwise you would need to ask me for access to push) and upstream is set to https://github.com/hello-robot/stretch_ros.git in order to sync local change (mainly updates to the robot description files based on our modification) while being able to get updates from the official repo.

To update stretch_ros:

cd catkin_ws/src/stretch_ros

# Fetch all the branches of the remote into remote-tracking branches
git fetch upstream

# Make sure that you're on your master branch: 
git checkout master

# Rewrite your master branch so that any commits of yours that 
# aren't already in upstream/master are replayed on top of that 
# other branch: 
git rebase upstream/master

Mapping

The main launch file is sp_navigation/launch/nav_3d.launch, the maps are stored in the same package, under the maps directory. The octomap package needs two maps, 2d map is stored in map_xxx.pgm and map_xxx.yaml, the 3d map is stored in octo_xxx.bt

2d mapping

  1. On the local machine, ssh into strech by typing ssh stretch or ssh [email protected] on the host computer.

  2. In the ssh terminal, run

roslaunch stretch_navigation  mapping.launch rviz:=false

and on the local machine, run rviz -d octomap_mapper.rviz

  1. Then use keyboard (i,j,l,, for forward left, right, and back) drive the robot around until the map is complete, then save the map by running on stretch:
rosrun map_server map_saver -f /home/hello-robot/catkin_ws/src/sterile_processing/sp_navigation/maps/map_<name>
  1. Modify the map_name argument in sp_navigation/launch/nav_2d.launch.

Octomapping

  1. On the stretch, run:
roslaunch sp_core robot_nav_bringup.launch
roslaunch sp_navigation octomapping.launch

  1. On the local machine, launch rviz, and use 2d Pose estimation option to give a reasonable starting point, so that the robot's current location align with the 2d map: rviz -d octomap_mapper.rviz

  2. Drive the robot around to finish mapping and save on stretch:

rosrun octomap_server octomap_saver -f /home/hello-robot/catkin_ws/src/sterile_processing/sp_navigation/maps/octo_<map_name>.bt
  1. Modify the map_name argument in sp_navigation/launch/nav_3d.launch.

Navigation

On stretch:

roslaunch sp_core run.launch

On local Ubuntu machine:

rviz -d navigation.rviz

Tips: To find out the map coordinates of a specific location, publish the point using rviz and see the terminal output. An example of the results: Position(0.23,0.808,0.000), Orientation(0,0,0.951,0.31)

Main state machine

When the navigation is launched, run:

rosrun sp_core state_machine_node

To start the workflow, you need to trigger a service call to /sp_sm/start. To post orders, publish a message to /sp_sm/post_orders (disabled now for demo) To clear orders, trigger call to /sp_sm/clear_orders

All three of these can be accessed through the web-based interface as well. (the order publishing has not been implemented yet, right now it just trigger the /sp_sm/start service )

There is also a gui for service call: rosrun rqt_service_caller rqt_service_caller

Calibration for the grasping

The states in the state machine are defined in src/custom_states.py. There are a few offsets that needs to be calibrated for better grasping:

  • Under MagnetState 
wrist_extension_offset_m
forward_offset_m
lift_offset_m
  • Under PostMagnetState 
post_grasp_lift_m
joint_wrist_yaw # radians first one: drop INCREASE a little
Clone this wiki locally