Fault-tolerant-trained MARL
Permanent communication loss evaluation
Duration: Permanent
Recovery mechanism: Fault-tolerant-trained MARL
Experiment definition
The experiment evaluates Fault-tolerant-trained MARL under a permanent communication loss injection. Fault onset is randomized between environment steps 40 and 60.
Expected evaluation record
| Metric | Recorded value |
|---|---|
| Degradation-area | 4.14 ± 1.60 (corrected; mean ± 95% CI, n=48) |
| Safety-violation excess | 12.9 ± 5.4 (weighted, control-subtracted; mean ± 95% CI, n=48) |
| Fraction of time healthy | 0.669 ± 0.012 (mean ± 95% CI, n=48) |
| Relapse count | 0.77 ± 0.18 (mean ± 95% CI, n=48) |
| False-recovery rate | n/a - no fault detector (trained policy only) |
| Intervention recall | n/a - no fault detector (trained policy only) |
Values are means with 95% confidence intervals across every target × duration-fixed combination shown on this page (see graphs below), 8 evaluation seeds each, for each of 3 training seeds. See the Results page for the cross-mechanism comparison.
Experimental trajectory graphs
One representative run (evaluation seed 0 of 8, training seed 1 of 3) per target × mode combination from the full sweep. See the Results page for the multi-seed aggregate comparison across mechanisms. The graphs for every seed are archived at doi:10.5281/zenodo.23304785 (files available from 10 November 2026).
Pressing Agent Target (blackout mode)
Pressing Agent Target (stale mode)
Notes
This page is intentionally separated from the paper-facing results so that full-resolution plots, additional seeds, and diagnostic figures can be documented without overloading the publication.