Recommendations for diverging trajectory optimization?

mikh3x4 / nerf-navigation

Code for the Nerf Navigation Paper. Implements a trajectory optimiser and state estimator which use NeRFs as an environment representation

https://mikh3x4.github.io/nerf-navigation/

MIT License

183 stars 24 forks source link

Recommendations for diverging trajectory optimization? #17

Open jfrausto7 opened 10 months ago

jfrausto7 commented 10 months ago

After experimenting a bit, I've noticed after running the simulate command some trajectories begin to gradually veer off further and further until they no longer seem to be reaching or optimizing for the goal position set in the script (see images below).

result Screenshot 2023-11-01 145553

I have found some (small) improvement by reducing the planner learning rate a bit. I was wondering if there are any tips on improving the stability of this trajectory optimization, or if this is a known behavior?

huevoypatata commented 1 month ago

I'm having the same issue, have you solved it?