Determine whether LP-BTS surpasses the strongest domain-engineered scheduler

Determine whether the LP-BTS learning-guided budgeted tree-search architecture produces a genuine survival improvement over the HQARRF-SC domain-engineered scheduler for one-to-many mobile charging under the central evaluation setting, rather than merely an unresolved observed difference.

Background

LP-BTS achieved a higher observed survival than HQARRF-SC on the sealed 30-scenario confirmatory bank, but the estimated paired survival advantage was +0.0066 with a 95% confidence interval of [-0.0037, +0.0184]. Because the interval includes zero, the evidence does not establish superiority over the strongest domain-engineered comparator.

The unresolved comparison is central to interpreting the practical value of explicit planning over the HQARRF-SC heuristic scheduler. The paper reports the result descriptively and leaves the question of whether the apparent advantage reflects a reproducible performance difference unresolved.