This may be a little out of bounds- but since the first Live Stream where flow fields were discussed Ive been really interested in the application of them in a real world robot, possibly even a mesh bot network. I'd love to get my hands on that testapp that is demoed in the Mar 22nd Live Stream and maybe explore it a little more. Is there any chance of that maybe being put out in Open Source domain - ie GitHub or something? Are there maybe some community projects already out there that you're aware of or could point me to? Thanks -ProC3ss
I think something like flowfield pathfinding would be pretty difficult (and may I add, pointless) in a real-world situation.
Unfortunately, I've already asked, here was the response: http://forums.uberent.com/forums/viewtopic.php?p=685593#p685593 Here are some articles that outline what I understand to be the basic idea of flowfield (even though they don't call it a flowfield): Collaborative Diffusion (link doesn't seem to be working right now but it'll probably be back) Using Potential Fields in a Real-Time Strategy Game Scenario The hard part about using these sorts of things is that you need to know your position (at least on a grid) in real-time, and know where obstacles are on that grid. In other words, the problem is not with the algorithm but with the robot's sensors. It's not undoable (in fact I kind of want to try it now ) but it won't be silky smooth like in PA or SupCom2. (Especially not SupCom2 unless you're using Mecanum wheels)
Thanks for the info Dallonf- See I was thinking precisely of using it in a scenario where I know my position on a grid- robotics/arduino/embedded linux is a hobby Im pretty passionate about. Ive built quite a few small bots and controlled them with everything from custom smartphone apps to onboard webservers - but Ive never really done an autonomous bot and its a project Ive had brewing in the back of my head for about a year now. All along the intention has been to generate a grid based on several things- floorplan / gps / motor feedback / distance sensor array / ir light sensor array. Using a database the bot would map out the grid and designate areas as either passable or impassible and relate that to gps positions. When I saw the demo app of the flow fields I was like- This is exactly what I want - ping on the distance sensor = grid cost, etc...
(to clarify) I don't think that the app demoed would magically drive a real world robot- but if I could take that app feed in my own obstacles or 'cost' and have it spit back out a path of coordinates as it relates to my own grid- then that would be rad.
Flow fields are pretty easy to code, took me 2 days to get one going for my game without any prior experience with them (though took a while to get it working exactly how I wanted and I am still tweaking it every now and again), all they really are is an array or two of costs, an array of vectors and function to determine in which direction the lowest cost tile is from among the 4 (8 if you include diagonal) cells around the one you are calculating the direction for, just have a try at coding it yourself.
