Longer distance

[tiralonghipol]

Hi! First of all, amazing work. Now, I set up a Gazebo world in which the target tree is far away from the initial position of the quad, and since it stops in the air after the first iteration (takeoff_forward_offset), I was wonder:

How can I increase the 'searching distance' ? And, moreover, is there any description of the used parameters (XY.yalm) ? Thanks again!

[chibi314]

Hello! Thank you for your comment.

There is no limitation of distance between the UAV and the first target tree (labeled by red cloth), so I think the cause is

• the limitation of URG range -> can't detect the tree
• the size of red cloth in the image -> can't detect the red cloth

I want to know this problem in more detail. So could you please give me more information by source code or screen shots?

Thanks!

[tiralonghipol]

More specific: my idea is to 'adapt' your algorithm to travel a tunnel until the red cloth is found. So, just to make a test, I build up something like this:

For now there are 2 related problem:

• the tunnel is too long, so if I keep the takeoff_forward_offset (inside motion.yalm) = 4, after some time the quad stops in the air (I think cause it never reaches the red cloth?)
• if I increase the takeoff_forward_offset (like 40, for example), the quad reaches the tree with the cloth but if flips immediately after

How should I set the parameters in order to reach the end of the tunnel (50m more or less) ? Thanks again.

Ps. If you prefer other communication channels write me at paolo@triesterivista.it

[chibi314]

That's a nice work!

Is the UAV the same as original one? It should have RGB camera and LRF.

So, please check if the tree is detected correctly. After the launch your program, show the "/target" as follows

rosrun rqt_image_view rqt_image_view topic:=/target

Then, start the UAV. Wait a minute, and if the tree is detected, you can get a image like this.

The blue boxes are detected trees by LRF. The area surrounded by the white contour is the cloth detected by camera. Please show me the screen shot of "/target" image.

[tiralonghipol]

Yes! I did not change the uav configuration

This is what I get if I place the tree close to the quad starting point:

If I start from the same point but the tree is placed at the end of the tunnel, the quad does not detect it. How can I tell the quad to keep searching until the tree is found? (with collision avoidance, of course)

[chibi314]

In our implementation, if the UAV can't detect the tree after taking off, the task stops. This is suitable for the challenge(http://www.lsse.kyutech.ac.jp/~sociorobo/ja/forestdrone17). This is Japanese, sorry.

How can I tell the quad to keep searching until the tree is found?

So, if you want to do this, you have to modify jsk_uav_forest_motion/script/motion.py.