0
$\begingroup$

In the Literature I have read so far, I saw RRT star was running in multiple loops to converge to a better solution ( near optimal). I was wondering how I could implement the same, as most pseudocodes don't explain that part.

$\endgroup$
0
$\begingroup$

RRT* indeed runs in multiple iterations to converge to an asymptotically-optimal solution.

In simple words: you could understand RRT* as an RRT, but with: 1) the stopping criteria being time/iterations instead of reaching the goal node and 2) with an extra re-wiring step that chooses, inside of a neighborhood of radius r from your already wired sample, the neighbor that can see its path cost more reduced, and attempts to connect to it. With this last methodology you get more styled branches which lead faster to the goal.

Here you got a video that explains RRT.

The same author makes another video on RRT* (in particular the re-wiring process).

$\endgroup$
  • $\begingroup$ How is the goal corresponded out of the explored nodes if that is the case? $\endgroup$ – Ajin2305 Aug 20 '18 at 11:52
  • $\begingroup$ @Ajin2305 the goal node has already been found, then you keep iterating to refine the branches. You do not lose the goal node you found, but might find a new path to it due to this refinement. Is that an answer to your comment? $\endgroup$ – daniglezad Aug 20 '18 at 16:03
  • $\begingroup$ I got the following characteristics for convergence, is it right? robotics.stackexchange.com/questions/16318/rrt-star-convergence $\endgroup$ – Ajin2305 Sep 3 '18 at 22:47

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.