Aditya's profile - activity

I am trying to install ROS on the latest version of gentoo and it seems there is a mismatch in the way ROS defines package dependencies. Specifically, rosinstall tries to satisfy dependencies by installing "dev-util/gtest" whereas gentoo maintains gtest at "dev-cpp/gtest". It would be very helpful if this issue is resolved soon with high priority. Thanks!

I launched a node with roslaunch. The node uses ROS_*_STREAM macros to log information. The log file created by the node however shows spurious characters. Following line shows an example

^[[0m[ INFO] [1325453858.376186830]: Alasca scan: time=-0.000000, points=31, nodes=11^[[0m

Every line in the log file starts and ends with "^[[0m". Does anyone know what might cause this behavior. This happens with not just this node, but all nodes that I launch with roslaunch.

ROS: Electric, OS: Ubuntu 11.10 x86

I am not sure if this is the right place for an OpenCV question, but I will ask nonetheless :) I want to compute distance transform on a cv::Mat object. However, for each pixel in the matrix, along with its distance to the nearest zero pixel, I want to identify the nearest zero pixel (row/col) as well. Is there any in-built OpenCV functionality to achieve this? Thanks!

While using "rosmake -i" for a package that depends on a couple of other packages, I realized that the "-i" flag, which creates ROS_NOBUILD file, operates only on the current package and not its dependencies. I think it is desirable to either modify behavior of "--mark-installed" option to mark the package and its dependencies as "installed" or introduce a new option, say "--mark-installed-recursive", that will do the job. Any comments are appreciated!

I am wrapping a third party library in ROS. The library has a few executables as well. I use a custom Makefile to build the library and the executables. However, when I do "rosrun customlib executable", I get "shared library not found" error. Should I modify the Makefile and use rpath into the executables or is there any default mechanism to take care of such cases? Thanks.

I am trying to find out best practice for using ROS diagnostics in hardware drivers. For example, say I am writing a driver to communicate with a device over serial port. When device node is launched, assume that the serial port could not be opened. In such a case, should the device node keep running with diagnostics updates or should it end the process? I would choose diagnostics updates over abort. Any comments are appreciated!

Assume stack A depends on ros stack and its stack.xml file has "<depend stack="ros"/> and assume that I create another stack B which depends on stack A and stack ros. If B's stack.xml specifies <depend stack="A"/>, does stack B also inherit "ros" stack as its dependency or do I need to explicitly specify <depend stack="ros"/> as well? Thanks.

I have successfully wrapped an external library into ROS framework following instructions in the tutorial. However, I have used a stack and a package to acheive this. I would like to switch to using a unary stack instead. I am not sure about how to create a unary stack. I do not use CMakeLists.txt in my package for wrapping the external library and as such CMakeLists.txt files in stack and package are not a problem. However, what should be done about the Makefile(s) in stack and package in order to create a unary stack from current stack-package. Thanks!

I enabled the ROS Electric repository in Ubuntu 11.10. However, while installing ros-electric-desktop-full, I ran into the following dependency conflict - ros-electric-vision-opencv depends on libopencv2.3-dev version 2.3.1+svn6514+branch23-7~oneiric; however, the version of libopencv2.3-dev available is 2.3.1+svn6514+branch23-8~oneiric. I think the ros-electric-vision-opencv dependency list needs to be updated to bring it in sync with the version of libopencv2.3-dev.

While installing ros-electric-desktop-full, I ran into a conflict of ros-electric-vision-opencv (through libopencv2.3-dev) with the default OpenCV in Ubuntu 11.10. In order to resolve the conflict, I manually installed libopencv2.3-dev, which removed Ubuntu's default OpenCV installation (libcv-dev, libcvaux-dev and libhighgui-dev). However, I found out that the FindOpenCV.cmake file that can be downloaded from the OpenCV website does not work with libopencv2.3-dev. It used to work with default Ubuntu OpenCV installation though. That FindOpenCV.cmake file looks for OpenCV libraries with very different names than those installed by libopencv2.3-dev. Is there a FindOpenCV.cmake file that works with 2.3.1 and may also be compatible with default Ubuntu installation (libcv, libcvaux, libhighgui)? This is definitely a show stopper.

Which encoding is most appropriate for using cv_bridge to convert cv::Mat_<cv::vec3d> matrices into ros images?

I successfully installed ros/electric in Ubuntu 11.04 (Natty) using

sudo apt-get install ros-electric-ros-base

However electric now uses eigen3 and it was not installed during electric installation (I earlier had eigen2 installed). Aptitude search shows following:

agadre@riverine:~$ aptitude search eigen

i libeigen2-dev - lightweight C++ template library for linear algebra
p libeigen2-doc - eigen2 API docmentation
p libeigen3-dev - lightweight C++ template library for linear algebra
p ros-electric-eigen - Eigen Library
p ros-unstable-eigen - Eigen Library

Why was ros-electric-eigen (and by extension libeigen3-dev) not installed? Currently default eigen in Ubuntu 11.04 is libeigen2-dev.

This same question also extends to opencv-2.3.

Thanks!

Is it possible to display a PCL point cloud of type pcl::PointCloud<T>, where T is a custom class, in rviz? Specifically, are there any requirements/limitations on the custom class T for it to play nice with rviz? It seems that the pcl_visualizer does provide a framework for displaying pcl::PointCloud<T> data through custom handlers. Thanks!

I am working on a similar problem. I have a dynamical model of a boat (Nomoto turn-rate model, with side-slip and linear speed dynamics added) that I am trying to use to generate motion primitives. As I understand, ideally I should choose primitives that end up in specific grid cells. However, as mentioned by Martin, it is not necessarily easily possible with the non-holonomic model I am using.

In my case, currently I am applying certain control actions for a specific period of time to forward propagate the dynamic model. Then whatever end-pose I end up with, I map it into its corresponding grid cell. Depending on discretization used, this approach will result in discontinuities at end grid points.

I was wondering if anyone else has come across such a scenario. I would certainly appreciate an input. Thank you very much for your help!