Point-Mass REDQ Baseline results
Created on December 17|Last edited on December 17
Comment
Sparse rewards
Step 5000:

Step 15k:

Step 30k

Step 100k

Step 5k

Step 15k:

Step 30k:

Distance-to-goal with tolerance
Reminder: this reward function is always positive, and goes from 0 (when distance \in (initial_distance, \infty)) to 1 (when distance \in [0, 0.05))
Step 5k:

Step 15k:

Step 30k:

Step 100k:

Add a comment