I have went through numerous papers recently. My search terms are starting to return the same papers all the time, but every now and then there are some good ones worth sharing. Here's some which are relevant to my recent prototyping results.
Very similar approach to directive circle method. It discusses the reasoning behind using only maximum velocity when selecting the velocity.
I had the gut feeling that multiple iterations would help the velocity selection. These papers provide some more information on the subject.
This paper discusses how VOs can be used in formation type of situations.
This paper discusses how the time horizon (how much the cone is clamped) can be adjusted.
Zvi Shiller was co-author of the well known velocity obstacle paper and has papers and videos at his page regarding the subject. Those non-linear velocity obstacles look pretty cool!
Don't suppose you've managed to get your hands on any papers on the non-linear velocity obstacles?
ReplyDeleteI'd love to see some source code reflecting either this or the recursive probabilistic velocity obstacles. Hell, I'd just love to see some source aside from RVOLib.
Just how to determine:
a) a generalised reciprocal velocity object
b) the time to intersection with it
seems to be beyond my tiny mind - and yet you seem to have done it overnight!
Tip of the hat to you.
My interest in velocity obstacles as silver bullet to solve all the local navigation needs is diminishing. But it is very interesting approach still. I'm looking forward what the UNC guys can come with next.
ReplyDeleteMany NLVO papers imply that they are using sampling/rasterization to make the implementation possible.
One thing I like about VOs is that they tell what velocity you should not choose, but they kinda still does not solve the problem that which velocity you should choose :)