Mike Gao's profile - activity

Hello! Recently I notice that, after I create a TransformListener object in a class in python, e.g.:

self.tf = TransformListener(),

the created tf object gets stuck in the memory, even I try deleting it in the "__del__(self)" method of the class with:

del self.tf

self.tf = None,

it does not work. The TransformListener object is used by a method of this class. The problem is that when I continue to create the object of this class (which is called by another class during operation), the TransformListener objects accumulate and occupy the computational resources. Thus my question is that how I can get rid of it to free the unnecessarily occupied resources? I appreciate any answers and suggestions in advance!

Hello! I have a question regarding the basic knowledge about ros (especially hydro), which is, when the ros program starts, how does it locate the needed resources, e.g., libs, executables, etc. ?

Currently I install my ros in /opt/ros/hydro, and in hydro it seems that the source codes are not installed together any more. So if I want to make some changes to some ros pre-installed packages, e.g., ros navigation packages, I have to clone the source codes of the related packages in another directory (usually in the catkin workspace), and make changes and build them to get new libs or executables (so there are two ros packages with the same name now), then how could I make the program use the newly built libs without copying (or installing) them to replace the old ones in the ros install directory?

Thanks for any help!

Hello, I would like to know, is there the implementation of the Vector Field Histogram(VFH) method for 2D mobile robot navigation in ROS?

I know ROS has implemented the Dynamic Window Approach(DWA) in the "navigation" stack, but it seems that this is not very suitable for the holonomic robot. Although the "Player" has the implementation of VFH, it cannot be used directly in ROS, so is there the ROS version of it? Or is there the implementation of other state of the art for local motion plan especially for holonomic robot in ROS? I would appreciate for any clues or suggestions, thanks!

Hello! I have two Sick LMS100 laser scanners on a mobile platform, each of which covers 270 degrees. I put them in diagonal direction on the robot to cover 360 degrees. In order to properly combine the data from the two laser scanners, my question is, is there implementation within ROS, which can decide the relative positions of the two laser scanners to each other? Currently, I can only manually tune the positional parameters to combine them. Further, how do you usually measure the extrinsic parameters of a laser scanner to the robot center, is there software tool like for the camera doing this work?

Thanks a lot for reading, and I will appreciate any feedback or suggestion from you!

Without spinOnce() function, you get nothing from ROS (no messages, no services, etc).

Hi, I would like to ask, if there is solution from hector group to localize the robot against the known map. As the hector_slam provides excellent localization, but it also builds new map at the same time, on the other hand, the amcl package of ros does not work well with my robot (a segway omni platform), especially compared to the hector_slam, so I try to directly use the hector_slam with the known map (by just ignoring the produced map and only use the localization results), but the map being built does not always match the known map, therefore the localization is not as expected. Thus I need a better solution for this (not only from hector). I use ros-hydro on ubuntu 13.04 32bit. Thanks a lot for viewing! (As far as I know, hector_localization package does not do this job.)

If I do not misunderstand, it seems that you could try setting the Gazebo simulator to be paused when it starts in the launch file, and you could start the simulation with the start button in the GUI of the simulator after you are sure that everything is correctly spawned.

Hi, recently I am learning tf package. I have read the tutorials, but have a small questions concerning the source code of transformlistener, which is: "vel_msg.angular = 4 * atan2(transform.getOrigin().y(), transform.getOrigin().x());" in the line 35. Why dose it need to multiply 4 here? For the aim of more quickly tracking? I know the scale value for keyboard control is 2, but it should be irrelevant to this, as this is calculated from the coordination. Appreciation for any help!

Hello! I am using ros hydro and my os is ubuntu 12.04 32bit. I want to know if there are suitable depth sensor filters in ros hydro which can filter out the points colliding with the robot self parts based on the robot model (e.g., through urdf file). Currently I am using a footprint filter for 2D laser scanner, and it seems to work fine, but I do know if there is such kind of filter for 3D sensor, e.g., kinect. And also in ros hydro, things are changing, so I would like just to ask. Thanks in advance for any suggestions and helps!

Hi,

When I use rosrun to begin a package, it works fine, but when I change to roslaunch it (it is a very simple launch file with only one node, and the package does not need any parameters), the package fails, the error information is: "process has died [pid 13287, exit code 255]." Nothing valuable is shown in the log file. Appreciation for any suggestions!

Hello! I encounter a really strange problem about ros::init(). I successfully compile the whole program, and of course the ros::init function is called at the first line of the "main program" as usual, then the creation of ros::NodeHandle is called. But when I run the program, the ros tells me that "You must call ros::init() before creating the first NodeHandle", then stops the program. And I also checked the related source code which says it is because the ros initiation fails (well, naturally), but why? Can anyone provide some clues? Really appreciate!