PR2/ROS/HRL Related Issues

From Healthcare Robotics Wiki
Revision as of 14:02, 20 June 2012 by Advaitjain (Talk | contribs) (Maintenance Log)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Please record all issues that you think others might encounter while using the PR2 in HRL.

Software

  • Commands to default joint controls execute instantaneously without using splines creating dangerous robot motions:
    • This occurs due to chrony weird-ness. Restarting chrony seems to fix the problem:
sudo /etc/init.d/chrony restart
  • We were unable to use the joystick to move the torso up (Advait, Jun 23 2010).
    • If this happens again then we need to report it to Willow with appropriate diagnostic information.
      • Check the diagnostics to see the effort on the torso joint. If it is zero, the torso won't move.
      • List all controllers and make sure nothing else is commanding the torso. (Only the torso_controller)
      • Stop the torso_controller using pr2_controller_manager and run: "rosrun robot_mechanism_controllers effort.py torso_lift_joint"
      • Restart your robot (sudo robo start) and retry everything.
    • Power cycling the entire robot fixed the problem for Advait.
  • Messages about /ground_object_cloud being late by n seconds where n increases with time.
    • Completely shutting down and power cycling the robot will fix this.

Disabling Arms Calibration

If the robot is ever shutdown where the arms are obstructed, when the PR2 is started up again it is necessary to disable the arms' calibration so that the robot can be moved away from the obstruction. If the arms' calibration is unable to complete due to the obstruction, the casters will never calibrate and the base will be unmovable. Without running "robot start", execute

roslaunch /etc/ros/diamondback/robot_no_arms.launch

to launch the robot and enable the runstop so that it can calibrate the casters. At this point, a teleop node can be run so the base can be moved. Kill the launch file and run "robot start" to complete the calibration for the arms.

Network

Communicating between the PR2 and other computers in the lab

  • To communicate with ROS while using gtpr2LAN, you must define ROS_IP in your .bashrc. One quick and dirty way is to insert:
export ROS_IP=`ifconfig $1 | grep "inet addr" | gawk -F: '{print $2}' | gawk '{print $1}' | grep "68"`
  • As of June 23, 2010 this should work without any modifications if your ROS_MASTER_URI is set to http://pr2c1.hsi.gatech.edu:11311
  • If it does not then
    • ping pr2c1 -- if this does not resolve to an IP try restarting network (sudo /etc/init.d/networking restart)
    • if things don't work even after a network restart, talk to Advait.
  • Information on the old HSI DHCP Bug.
GTwireless login information

For the robot to have a wireless connection to the internet, it must be logged into LAWN. To log the robot into the network, go to the LAWN device login and use the following MAC address to login with this your username and password. Make sure that ISS is DISABLED and login a different device is ENABLED.

  • Ask an HRL member for the MAC address of the robot.
Troubleshooting
  • Launch talker/listener on the machines that are having trouble communicating. Look at the hostname field given by rosnode info <your_node_name> and make sure that the two computers can talk to each other using the hostname printed out with rosnode.
    • To find out your node name type
rosnode list | grep <some_part_of_your_node_name>

Internet and or graphics card not working on base station

  • on June 6, 2012 Advait and Hai saw this problem. They debugged it down to a newer kernel that probably got installed when upgrading to Electric.
  • To fix this:
    1. open /boot/grub/grub.cfg
    2. change the default menu item to be an older version of the kernel.
    3. on June 6, the older verison was 2.6.32-41 and a 3.0.6 kernel was the one causing problems.
    4. reboot.

Sometimes finicky hardware?

  • around mid-Jan 2011, all the joints were reporting stale ethercat packets. Power cycling the robot (and not simply rebooting) fixed the problem. -- Advait
  • Feb 14, 2011: Ethercat timeouts on the l_gripper_motor and l_wrist_r_motor are causing breaker halts -- Travis
    • Ethercat timeouts tend to correlated with gripper making contact with the world.

Maintenance

Instructions

  • Before performing any maintenance on the PR2, please look at the following:
  • Look at the specific procedure for what you would like to do. The complete list is here
  • Use only the tools provided with the PR2 and put them back in the PR2 toolkit after use.
  • After performing the procedure, make a note in the maintenance log.

Verification Steps

Here are some helpful scripts to run after you perform maintenance on the PR2:

  • After replacing the PPS sensors and/or covers, on the base station, run:
cd /opt/ros/cturtle/stacks/pr2_ethercat_drivers/fingertip_pressure/scripts
./view_fingertip_pressure board:=/pressure/l_gripper_motor  #change l to r to test the right gripper's sensors

Maintenance Log

  • July 5, 2010. Replaced the fan filter, cleaned the original filter and put it back with the rest. (Advait)
  • July 13, 2010. Front bottom panel was not properly tucked under the rear bottom panel. Followed page two of this. (Advait)
  • August 12, 2010. Replaced fan filter (Hai)
  • August 13, 2010. Replace PPS sensor cover on a finger of the right gripper (Hai & Advait).
  • Nov 12, 2010. Replaced the four powered casters on the base (Marc and Advait) -- Ethercat packet dropout errors, possibly faulty ethernet cables in the casters, Willow sent us four replacement casters.
  • Nov 12, 2010. used small vacuum to clean the air filter. (Hai, Marc and Advait)
  • Dec 2, 2010. Replaced PPS sensor covers on all fingertips of the both the grippers (Tiffany & Advait).
  • Jan 28, 2011. Cleaned the fan filter by washing. (Tucker)
  • June 16, 2011. Cleaned fan filter with vacuum. (Tucker)
  • June 16, 2011. Replaced PR2 Run-Stop batteries. (Tucker)
  • June 01, 2012. Disconnected and reconnected front left caster ethercat cable. (Tucker and Kelsey)
  • June 15, 2012. Replaced the Front Left caster (Advait, Hai, Tucker)
  • June 19, 2012. Replaced the rear power panel. Hai had noticed that the power cable was damaged and willow sent usb replacement cables and panel. (Advait and Survy)
  • June 19, 2012. Replaced the 16 port network switch. Seems to have fixed trouble with cameras dropping packets and c2 being really slow. (Advait and Tucker)

Software Configuration Log

Hardware

  • Any 20A extension cord (with the appropriate connector) will work with the PR2. e.g. this