Lexicographic Constraint Optimization
in Hotel Management
Dynamic Online Control, Reinforcement Learning Adaptation, and Multi-Property Cloud/Quantum Integration
Antonios Valamontes — Kapodistrian Academy of Science, Tampa, Florida
We present a comprehensive framework for applying Lexicographic Constraint Optimization (LCO) to multi-tier hotel management. The system enforces strict precedence of safety, accessibility, and legal obligations over economic and operational objectives, embedding explicit ethical governance within the optimization hierarchy. Both static and dynamic formulations are developed, extending LCO to online stochastic arrivals through rolling-horizon re-optimization.
VIP Walk Rate
0.0%
Zero VIP displacement guaranteed
Overbooking Reduction
55%
OB slack 4.0 → 1.8
Revenue Preserved
98.2%
Only 1.8% trade-off vs scalar
HK Turns Reduced
14%
28 → 24 turns
Overtime Reduced
25%
16 → 12 shift-hours
Energy Savings
9%
100 → 91 proxy units
6-Tier LCO Hierarchy
Click any tier to explore — Figure 1 from the paper
Scalar Baseline vs LCO
Figure 4 — Normalized to [0,1], higher is better. †Inverted scores (1 − normalized raw)
- Scalar Baseline
- LCO
Table 2: Synthetic Validation (10×5 Instance)
LCO preserves Tier L₁ feasibility in all runs, trades < 2% revenue for large reductions in risk and operations.
| Metric | Scalar Baseline | LCO | Improvement |
|---|---|---|---|
| Rev (normalized) | 1.000 | 0.982 | −1.8% |
| VIP Walks (%) | 2.5 | 0.0 | — |
| OB (sum wₐ) | 4.0 | 1.8 | 55% |
| HK (turns) | 28 | 24 | 14% |
| OT (shift-hours) | 16 | 12 | 25% |
| EN (proxy units) | 100 | 91 | 9% |