kabamaru's profile - activity

well thanks a lot i've figure it out, after all .

I have to say that my package compiles ok. But I've seen people create a different include folder where they place their include file. I find this cleaner and a better practice in general and I want to follow it. The problem is I don't know how to do it without absolute reference. I don't want to use something like that:

#include "../../include/lib/foo.hpp"

my package name lets say it is upm. I want something like

#include "upm/include/lib/foo.hpp"

or without upm even better.

I've tried http://answers.ros.org/question/30425/include-directory-for-other-packages/

but with no success. Any suggestions? And generally how do you treat your header files? What it is considered good practice? Sorry if it's naive and a bit general.

I don't want new objects. I want a reference or else it doesn't worth the loss in speed. I will go with passing around the smart pointer. Thanks anyway.

I have a class that gets an sensor_msgs::Image::ConstPtr and there are a lot of member functions acting to this same image. void extractData(const sensor_msgs::Image::ConstPtr &image)

I was wondering if it is possible to have a data member instead of sending the pointer from one function to the other.

I've tried adding: sensor_msgs::Image::ConstPtr ℑ

But it says uninitialized reference member. Tried also removing Const or Ptr in declaration with no success. Any ideas?

Like shade said, since your angular velocity seems to be constantly 0, you should limit your angular range to something significantly smaller than a whole range scan.

The real problem of you approach is that you scan point by point and decide for each one. This seems completely wrong. This way your robot will decide for EVERY point in your laser scan (probably changing it's mind a lot of times within one scan). What could be better is for each scan first loop through all points (with a single for) and if there is a point that is below your threshold (eg 0.8 meters) then decelerate. Also for lowering the speed gradually you might consider using vel=vel*0.8 or something similar.

what type is buffer_image_compressed? sensor_msgs::CompressedImage?

Just for future reference. The best package if you have matlab:

http://www-personal.acfr.usyd.edu.au/akas9185/AutoCalib/index.html

By A. Kassir.

well it was a matter of some transform manipulation after all. the POVs of the camera and the laser (hokuyo) are not aligned. just for reference the tf that worked for me was: xyz ypr [0 0 0 -1.57 0 -1.57]

<node pkg="tf" type="static_transform_publisher" name="world_to_camera" args=" 0 0 1 -1.57 0 -1.57 /world /camera 100"/>

i've tried [ /world -> /camera , /camera -> /laser] & [ /world -> /camera , /world -> /laser] is this what you mean? in the second case they are both direct children of the /world frame.

I have a setup with a simple VGA Camera and a Hokuyo 2D Laser and I want to OVERLAY the readings from the Hokuyo on the window of the camera in rviz. I can see the hokuyo readings on the main window and I can also see the camera image in a separate window. I've already set the frame_id for camera to /camera and for the hokuyo to /laser. Also I publish the tf for /world frame to each of the aforementioned frames. The camera is next to the laser so the translation transform for the camera is 0,0,0 and for the laser is like 0, 0.05 ,0 which means 5 cm at the right. I want to see a line on the camera window much as like in the rviz tutorials (second video around 2:40).