Effort position controller fix #33
Conversation
| @@ -122,7 +122,15 @@ void JointPositionController::setCommand(double cmd) | |||
|
|
|||
| void JointPositionController::starting(const ros::Time& time) | |||
| { | |||
| command_.initRT(joint_.getPosition()); | |||
| ROS_DEBUG_STREAM_NAMED("joint_position_controller","Starting '" << joint_.getName() | |||
There was a problem hiding this comment.
I'd rather not make log statements in code that will be executed in a realtime context. Rosconsole macros end up making system calls which trigger a switch from primary (realtime) to secondary (linux) mode.
There was a problem hiding this comment.
I'll happily remove this but isn't it ok for the constructor and this starting function to be non-realtime?
There was a problem hiding this comment.
In the controller manager, starting and stopping a controller is done inside the realtime update() method. Controller construction and init()ialization is non-realtime ,though.
I just checked that the ControllerBase class documentation is not using the \name blocks present in other parts of ros_control. That, at least, can be improved:
/** \name Real-Time Safe Functions
*\{*/
...
/*\}*/
/** \name Non Real-Time Safe Functions
*\{*/
...
/*\}*/There was a problem hiding this comment.
Improved ControllerBase documentation [here]((pal-robotics/ros_control@4e23e2c).
There was a problem hiding this comment.
Removed the debug output.
Effort position controller fix
diff_drive: use backslashes in doxygen documentation
Create linear/quadratic meas cov model classes
update()andstarting()functionsThis fixes a bug where if you start the controller and your current joint position is slightly outside the joint limits, the starting commanded position is also outside the limit, and the
shortest_angular_distance_with_limitsfunction returns an incorrect value of 2*PI instead of a small decimal. This causes the error to be very high and causes a lot of noise in the PID controller.This bug is also issued here: ros/angles#2