Nowadays,there are some difficulties,so please,everybody,could you give me some helpful advice or some person interested robot path planning clues? I have focused on AUV(artificial underwater vehicle) path planning for couple of yeas,last year I useed genetic algorithm for solving 2D auv path planning problem and had good result( envirnment model:100*100 grids,search method:genetic algorithm). But now,I must use 3D envirment(meanwhile consider the influence of the ocean current),so this is possible use GA and 3D grids to overcome the difficulties. thank you in advance!

Assuming that your project is a simulation of a real world environment and that the path needs to be calculated before the AUV begins its task then GAs are a good technique. You don''t need to use grids either, you can use vectors which will give you a far superior resolution.

If you need to do path finding on the fly then GAs aren''t usually a good idea. Use something like A*.

If you need further assistance just ask.



Stimulate

quote: Original post by fup
If you need to do path finding on the fly then GAs aren't usually a good idea. Use something like A*.

Better yet, wait about 6 months until my PhD thesis is back from examination and is published and you can get the answer from that.

As a rough answer, check out the following paper:

@article{Williams01,	author = "Williams, M., and Jones, D. I.",	title = "A Rapid Method for Planning Path in Three Dimensions for a Small Aerial Robot",	journal = "Robotica",	year = 2001,	volume = 19,	pages = "125-135"}  

I don't particularly like this method as it has difficulties in non-linear fields...

Personally, I would suggest A* as a first iteration. You will need to expand nodes in the action space, rather than the location space. It will have problems though in a non-linear current as many nodes in the action tree will be co-located in the location tree... search time will be high... you can get around this by integrating motion for a set distance, rather than a set time.

If you want to be able to perform replanning during execution of the plan, then I highly recommend that you look for my thesis later this year. It deals with the problem of Autonomous Flight Management (online, real-time planning and replanning) for a robotic aircraft working within cyclones (hurricances), where the environment is highly non-linear and dynamic... about the worst scenario you can get!

Cheers,

Timkin

[edited by - Timkin on April 6, 2002 11:12:33 PM]