search

vita-epfl / CrowdNav

[ICRA19] Crowd-aware Robot Navigation with Attention-based Deep Reinforcement Learning
MIT License
560 stars 166 forks source link

Tuining Paramters #56

Open sunilsrivatsav opened 1 year ago

sunilsrivatsav commented 1 year ago

How did you tune or decide the RL parameters? Is there any way to tune parameters or heuristically decided?

ChanganVR commented 1 year ago

It's been a while since I did the project but the parameter tuning was nothing special. Just the common techniques you would do for RL training. If you check the log printing statements, you can find the useful parameters to watch for or tune. Value heatmap is also another way to visualize the network's output.