felix's profile - activity

In simulation, my experience is that time/clock is managed by the simulator and you get a lot of overhead because of thi

I think I see where my problem is coming from. I used the robot frame as "fixed frame", so in this frame, odometry diffe

I think I see where my problem is coming from. I used the robot frame as "fixed frame", so in this frame, odometry diffe

I think I see where my problem is coming from. I used the robot frame as "fixed frame", so in this frame, odometry diffe

rviz odometry plugin does not respect keep value It seems that if the frame in which the odometry is displayed is the sa

I was trying to get ahead of my colleagues and port my OpenPRS/PR2 library under Diamondback, but it looks like the PR2 stacks and actions have not yet been released under Diamondback. Can somebody confirm?

I did the Desktop-Full Install: (Recommended)

sudo apt-get install ros-diamondback-desktop-full

and include such as pr2_common_action_msgs/TuckArmsAction.h are nowhere to be found.

cancelGoal crashes if called before one goal has been sent.

OpenPRS comes with ABSOLUTELY NO WARRANTY.
[ERROR] [1298889427.632366743]: Trying to cancelGoal() when no goal is running. You are incorrectly using SimpleActionClient
[ERROR] [1298889427.632673206]: Trying to cancel() on an inactive ClientGoalHandle. You are incorrectly using a ClientGoalHandle
Segmentation fault

if you send at least one goal, then it just reports the problem but keep running.

[ERROR] [1298889509.884434617]: BUG: Got a transition to CommState [PREEMPTING] when in SimpleGoalState [DONE]

Is this normal? should I fill a ticket?

(I know, it is stupid to cancel something which does not yet exist... but it should not crash.)

Just to add that cancelAllGoals behaves better and does not crash. It does not even report any error.

By the way, where are the source for these functions (cancelGoal and cancelAllGoals), I cannot find them in my ubuntu virtual box appliance.

The MoveArm action returns an error_code in its result:

more MoveArmResult.msg
#An error code reflecting what went wrong
motion_planning_msgs/ArmNavigationErrorCodes error_code

which I can access then with:

result->error_code.val;

is there a way to get a string (which is more meaningful than an int) from this error code? something like the ENUM defined here:

/opt/ros/cturtle/stacks/motion_planning_common/motion_planning_msgs/msg/ArmNavigationErrorCodes.msg

# overall behavior
int32 PLANNING_FAILED=-1
int32 SUCCESS=1
int32 TIMED_OUT=-2

# start state errors
int32 START_STATE_IN_COLLISION=-3
int32 START_STATE_VIOLATES_PATH_CONSTRAINTS=-4

Thanks in advance,

I find the notion of action server "limited" to one goal rather restrictive... Is it possible to have a process with more than one action server?

There is a difference between limitation to one goal and limitation to one action server.

• There is no limit to how many action servers a node can have. It is perfectly possible to have one ROS node with multiple action servers; just pass them different topic names on creation.
• Each SimpleActionServer is limited to one goal; whenever there is a new incoming goal, the currently pending goal is cancelled. This behaviour is a simplification of the actionlib API which catches most use cases.
• If you want to have a more sophisticated handling of goals, you are free to do so using the actionlib API directly (without the SimpleActionServer). You could have multiple goals executing in parallel, queue goals, and generally do whatever you want.

Just a side remark: if you have separate functionalities which you would ordinarily implement as separate ROS nodes, but which have to run in the same process for technical reasons, you could consider implementing them as nodelets.

I find the notion of action server "limited" to one goal rather restrictive... Is it possible to have a process with more than one action server?

For example, imagine that your have a process handling a particular "functionality" on your robot, and you want this process to handle different types of action goals (which need to execute in the same memory space). Could it be possible to define multiple action server in this process, or there will be some problems (such as reentrance, etc).

Of course, you could define a "meta" goal and use one of its arguments to define which underlying action/goal to execute.

Thanks found them.

Just a side question... any particular reason to get WG PR2 Desktop for non WG people? I could not figure out what are the additional packages/stacks in it.

Please follow the PR2-specific instructions for installing Diamondback:

http://www.ros.org/wiki/Robots/PR2/diamondback

Installation page for specific robots are linked off the ROS installation page:

http://www.ros.org/wiki/ROS/Installation

(5 robots and counting).

If you're really curious as to what a particular variant contains, you should look at REP 108 for the common entry point descriptions. If you're really, really curious, you can look at the rosinstall file for that variant:

http://packages.ros.org/cgi-bin/gen_rosinstall.py?rosdistro=diamondback&variant=desktop-full&overlay=no