Request for the 30 seeds / model artifacts used in Fig. 4 (IROS 2023)

[yueqiu666]

Hi, thank you for the code and pretrained models for "Robust USV Navigation
with Distributional Reinforcement Learning" (arXiv:2307.16240).

I am trying to faithfully reproduce Fig. 4. The public pretrained_models/
directory currently only contains seed 3 (IQN/seed_3 and DQN/seed_3). However,
Fig. 4 appears to report mean +/- std over 30 seeds, and I could not find the
30 numerical seeds or the 30 model artifacts in the repository history,
branches, tags, or releases.

Could you please share any of the following?

1. The list of the 30 training/evaluation seeds used for Fig. 4.
2. Whether the 30-seed result is the mean over 30 independently trained models,
   or one model evaluated with 30 seeds.
3. The 30 pretrained model checkpoints, if still available (otherwise just the
   seeds are enough for us to retrain).
4. Confirmation of the evaluation protocol (episodes per scenario, evaluation
   seeds, and how success/time/energy metrics were aggregated).

Thank you very much for your help.