ParNurZeal's profile - activity

In navigation stack, the tutorial is based on a static map and amcl for localization. When using with slam, some people suggested just exclude the static map and amcl and run gmapping.

I already did that and it worked. However, when I looked into rxgraph, why the only topic move_base(Navigation stack) use from Gmapping is /tf and it does not use /map from Gmapping at all?

I've looked into the code and guessed it might create map by itself in costmap2D from laser scan in move_base node.

If this is true, would it be better to use the real map from Gmapping?

Only costmap from laser scan and correct position from Gmapping through /tf is enough correct?

I am using P3DX, RosAria, Hokuyo URG-04LX-F01 to collect data as a bag file and put it into gmapping.

The robot moved very slow (0.03m/s). But I wonder why the map created from gmapping wasn't clear. It seems many data disappear along the way. And when I saw the process while gmapping was working. The registered data was very little. For example, after gmapping registered first scan, the bag file went for around 10 secs then gmapping registered a new scan again. Maybe because of this some data is disappeared.

I attached map image Gmapping created as well. http://www.flickr.com/photos/parnurzeal/6363494549/

How can I fix this? what is the cause for this?

Hello,

I used P3DX with Hokuyo URG-04LX limiting range to only 3 metres and gave the odometry data to Gmapping as well. I have problem in using Gmapping in long corridor where the feature look the same. For example, the left and the right side of robot is just the straight line. When I do Gmapping, even it walk through these parallel straight line for long time, it think it's in the same place and not move at all in gmapping in rviz. The map will update when it found something like doors. Thus, the robot walk for about 5-6 metres long until finding the doors but the map is 1-2 metres long.

How to fix this? or It's hard to do slam in the long corridor without distinct feature?

I am trying to do Gmapping on P3DX using Hokuyo sensor and using RosAria in controlling.

I have seen many suggestion on doing TF like following tutorial.

http://www.ist.tugraz.at/robotics/bin/view/Main/Poineer_mapping

In this tutorial, the code that transform odometry is below.

tf::Transform txOdom(odomQ, tf::Point((double) pos.getX()/1000.0, (double) pos.getY()/1000.0, pos.getTh())); tf.sendTransform(tf::StampedTransform(txOdom, ros::Time::now() ,"odom", "base_footprint"));

And the discussion below which one gave the code of tf from odometry to base divide by 1000 as well.

http://ros-users.122217.n3.nabble.com/Pioneer3-td1265101.html

They all divided by 1000 in position from odometry. Why?

The problem I have when I divided by 1000 is gmapping seems to receive only first scan and do nothing else and wait for something. I guess Gmapping think the robot doesn't move. So, I tried to delete 1000 out and now Gmapping accept more data and make a map. I am not sure that I do it right or not in deleting 1000 out.

Could anyone explain the meaning of this? I would be very appreciated. like what is the unit from aria and RosAria and what is the unit gmapping want to receive?

This might not involve with technical issue. However, I have a lot of problem in getting in the ROS website may be 3 times already in this month. Is it just me? or everyone encounter this problem?

I ask this because if it's just me, I then can find the cause and solve the problem myself.

I know that there is already the question about this in How to set unknown pixel as goal in move_base. That question's owner successfully did it by not using static map. However, that means the robot will not use the updated map from Gmapping in navigating, it will use rolling windows as local_planner use. This is not the answer I want. I want the robot be able to really navigate to unknown cell with the map from Gmapping.

Some has suggested to set "navfn/allow_unknown" to true. I have already set it but the robot is still unable to compute a path to the unknown cell goal. I would be very appreciate if someone know how to do this. Thank you!