robustfieldautonomylab / distributional_rl_navigation Goto Github PK

Dear author,

Thank you for sharing the excellent work!!!

I'm trying to run the "run_experiment.py" to evaluate the pre-trained models.

I'm wondering how can I visualize the evaluation of the results from "run_experiment.py" like you did in the README.md. I will be grateful if you could share the procedures.

Thank you very much.

Hello, I saw your work. It's really a great and complete job, but I have a question. Please tell me how to implement path planning for multiple unmanned ships based on your code.

Hello, Thank you for your work. I encountered an issue while running the run experiments file. Without modifying any settings, the experimental results for the complex environment [10obs, 8cores] are as follows:

=== Finish 500 experiments ===
adaptive_IQN | success rate: 0.71 | out of area rate: 0.12 | avg_time: 37.75 | avg_energy: 97.43 | avg_compute_t: 0.0010611089374018853
IQN_0.25 | success rate: 0.49 | out of area rate: 0.41 | avg_time: 38.70 | avg_energy: 99.71 | avg_compute_t: 0.001004070807167153
IQN_0.5 | success rate: 0.66 | out of area rate: 0.24 | avg_time: 37.52 | avg_energy: 94.62 | avg_compute_t: 0.001005096895758446
IQN_0.75 | success rate: 0.70 | out of area rate: 0.15 | avg_time: 38.17 | avg_energy: 97.55 | avg_compute_t: 0.001001747665702137
IQN_1.0 | success rate: 0.75 | out of area rate: 0.07 | avg_time: 37.78 | avg_energy: 97.41 | avg_compute_t: 0.0009959756102798688
DQN | success rate: 0.61 | out of area rate: 0.00 | avg_time: 33.05 | avg_energy: 96.16 | avg_compute_t: 0.0001960611697107753
APF | success rate: 0.58 | out of area rate: 0.07 | avg_time: 53.13 | avg_energy: 162.81 | avg_compute_t: 4.319793696453127e-05
BA | success rate: 0.37 | out of area rate: 0.01 | avg_time: 49.67 | avg_energy: 132.89 | avg_compute_t: 3.5731663451286436e-05

This doesn't seem to meet our expectations. Could you please let me know if there is any problem? :)