Limits of retrospective robot-task duration calibration and causal dispatch inference in open collective robotic construction logs
Exploratory audit of 70 distinct seeded weighted MaxCut graphs (8–16 vertices). Brute-force and MILP reference optima agreed in 70/70. Ideal state-vector QAOA achieved 64/70 optima at p=1 and 61/70 at p=2; simulated annealing and tabu each achieved 65/70 under 256 final candidates. QAOA training required 7980 exact full-state expectation evaluations, an additional unequal cost. These are CPU simulation results, not evidence of quantum advantage; no QPU or independent blind validation.
Authors
- Dr Starck
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-15
- DOI
- https://doi.org/10.5281/zenodo.24262785
- Primary Topic
- Quantum Computing Algorithms and Architecture
- Type
- preprint