sequenceDiagram
autonumber
box rgb(240, 245, 255) Internal Simulation Loop (High Fidelity Sandbox)
participant Orch as DayOrchestrator Dispatcher
participant Tourist as Tourist Thread (TouristTwin)
participant Staff as Housekeeper Thread (StaffTwin)
participant Geo as Geofence Validator Module
end
box rgb(255, 245, 240) External Target System (Hotel Platform)
participant API as Live Tourism ERP API
end
Orch->>Tourist: Initialize consumption time step (Time 13:00)
activate Tourist
Orch->>Staff: Initialize assignment time step (Time 13:00)
activate Staff
Tourist->>Tourist: Intent: Spontaneous room service food order from restaurant
Tourist->>Geo: Request tourist's current GPS coordinates
activate Geo
Geo-->>Tourist: Location: Pool loungers zone (Valid)
deactivate Geo
Tourist->>API: HTTP POST /api/v1/order/room-service (DTO: Suite 302, SKU Class-A Item)
activate API
API-->>Tourist: HTTP 202 Accepted (Order accepted by kitchen, inventory deducted)
deactivate API
deactivate Tourist
Staff->>Staff: Intent: Standard housekeeping of checked-out Suite 302
Staff->>Geo: Request housekeeper's current coordinates
activate Geo
Geo-->>Staff: Location: Residential building, 3rd Floor, Hallway (Valid)
deactivate Geo
alt Routine Step: Deducting Consumables from Floor Warehouse (ROUTINE)
Staff->>API: HTTP POST /api/v1/wms/chemistry/spend (DTO: Room 302, SKU Class-B)
activate API
API-->>Staff: HTTP 200 OK (Class-B consumables deducted from hotel balance)
deactivate API
else Staff Anomaly: Fraudulent Asset Write-Off as Defective (Forensic Fraud)
Staff->>Staff: Operational fatigue and cognitive sabotage increase
Staff->>API: HTTP POST /api/v1/wms/inventory/write-off (DTO: Bathrobe SKU Class-C, Status: Damaged)
activate API
API-->>Staff: HTTP 201 Created (Damage report registered, resource written off balance sheet)
deactivate API
end
deactivate Staff
Architectural Design of Autonomous Stochastic Simulators in the Tourism Industry
Designing High-Fidelity Sandbox Environments, Multi-Resource ERP Matrices, and Hierarchical Geo-Polygons for IT Footprint Generation
1 Introduction: The Paradigm of High-Fidelity Simulation of Tourism Facilities
This paper concludes the description of autonomous stochastic engines and details the architectural implementation of a high-precision simulator (High-Fidelity Simulation) deployed within the digital ecosystem of a tourism facility (resort complex, hotel, recreational zone). Unlike the first two papers, which focused on isolated domestic scenarios (FoodLifeCycleApp) and mobile workstations for manual laborers in agriculture (AgroLaborApp), the current study shifts to modeling a closed hybrid ecosystem where the streams of external clients (B2C segment) and the activities of service personnel (B2B segment) concurrently intersect.
The engineering priority of this third paper shifts from transaction generation speed to achieving the maximum depth and realism of the IT footprint. Under this paradigm, the simulator engine intentionally sacrifices performance: emulating a single operational week can take, for example, from 2 to 5 hours of real CPU time. The discretization scale of the macro-orchestrator narrows down to 1 second of simulated time, allowing for step-by-step calculations of personal agent schedules, honest handling of asynchronous responses from external APIs, processing of real network latencies, and simulation of cascading multi-resource transactions.
The generated data arrays serve as a reference baseline for training Process Mining systems (concurrent interaction graphs) and Forensic Analysis modules (identifying complex internal asset fraud within an enterprise).
2 Multidimensional Classification and Resource Circulation (ERP Master Data)
To accurately recreate the transactional workload on the warehouse management (WMS) and supply chain (ERP) perimeters of a tourism facility, the simulation engine utilizes multi-resource models. A comprehensive matrix operating with four invariant classes of inventory items (inventory / SKUs) is embedded directly into the In-Memory core:
- Resource 1 (Food and Beverages): Perishable items characterized by an expiration date parameter (
EXPIRED). They are consumed by Digital Tourists in the resort’s restaurants or via Room Service, and are also written off by kitchen Staff during meal preparation. - Resource 2 (Household Chemicals and Consumables): Detergents, professional disinfectants, laundry capsules, shampoos, and single-use toiletries. Used strictly by Staff (e.g., housekeepers) when executing cleaning work orders for the room stock. These represent the primary simulation zone for operational employee fraud.
- Resource 3 (Linens): Bedding sets, bathrobes, towels, and restaurant table linens. This class features a closed circulation cycle unique to the simulator:
Clean in central warehouse\(\to\)Issued to room\(\to\)Dirty/Removed\(\to\)Transferred to laundry (Write-off/Block)\(\to\)Returned to warehouse (Restock). - Resource 4 (Fixed Assets / Equipment): Electronic key cards, climate control remotes, decor items, and premium tableware. These possess a high stochastic (probabilistic) risk of loss, damage, or theft by both clients and employees.
3 Hierarchical Location Geo-Modeling and Spatial Discontinuities
To guarantee high-fidelity replication of physical behavior, the simulator’s Geofence Validator subsystem utilizes a three-dimensional, hierarchical spatial polygon matrix. The location of every building facility and agent is mapped to a tree-like spatial structural perimeter:
\[\text{Resort Territory} \longrightarrow \text{Residential Building} \longrightarrow \text{Floor} \longrightarrow \text{Room / Guest Suite}\]
Every agent action transmitted to the external API includes their current GPS coordinates. Unlike the agricultural simulator, where crossing a field boundary during an active session was categorized strictly as an anomaly, the hospitality engine implements the concept of legal spatial discontinuities.
A Digital Tourist can trigger a food delivery intent from the resort restaurant to their suite while physically sitting on a sun lounger by the outdoor pool. In this scenario, the Geofence Validator captures two distinct, valid, yet geographically separated polygons: GPS_Client (Recreation Zone) and Delivery_Target (Suite 302, Residential Building). The target system’s backend registers this transaction and fires off linked process tokens to coordinate the kitchen staff and delivery courier.
4 Architecture and Specifics of the Two Digital Twin Groups
The simulator’s internal core concurrently processes the states of two polarized groups of agents whose competitive interaction over the same spatial and resource footprint generates a unique digital footprint.
4.1 Group 1: Digital Tourists (B2C Agents)
The behavior of client agents is chaotic and emotionally driven. Their state vector is governed by subjective comfort factors, travel-induced fatigue, and compulsive impulses:
\[\mathbf{DT}_{\text{tourist}, i} = \langle \mathbf{A}_{\text{tourism}}, I_{\text{fomo}}, M_{\text{hunger}}, H_{\text{comfort}} \rangle\]
Where \(\mathbf{A}_{\text{tourism}}\) is the tourist preference matrix that clusters agents into specific profiles (e.g., Adventure/Extreme, VIP/Lounge, Budget Family Tour). The \(I_{\text{fomo}}\) parameter regulates the agent’s susceptibility to internal push notifications regarding ongoing resort events and activities.
4.2 Group 2: Digital Staff (B2B Agents)
The behavior of employee agents (housekeepers, front desk administrators, couriers) is strictly regulated by standard operating procedures and RBAC1 permissions. Their behavioral vector incorporates operational degraders:
\[\mathbf{DT}_{\text{staff}, j} = \langle \mathbf{A}_{\text{duty}}, E_{\text{fatigue}}, K_{\text{kpi}}, H_{\text{anxiety}} \rangle\]
Where \(\mathbf{A}_{\text{duty}}\) represents the professional assignment matrix, mapping the agent to specific floors or zones within the facility, while \(E_{\text{fatigue}} 'in [0.0 \dots 1.0]\) is the cumulative exhaustion level that directly reduces procedural compliance quality.
5 Sequence Diagram: Interaction Between Tourists and Staff
To visualize an end-to-end simulation time step, a parallel process execution model has been designed. Within the High-Fidelity framework, agent threads simulating tourists and staff concurrently interact with a shared room stock and different resource classes (e.g., Food and Household Chemicals) while passing coordinate spatial validation checks.
WarningImportant Note
This public documentation does not cover all steps and scenarios for the application specifically, or for the digital business process simulation ecosystem as a whole.
6 Decomposition of Forensic Anomaly Classes and Fraud Patterns
Because the high-fidelity step-by-step simulation integrates inventory and warehouse accounting logic, under the combined influence of fatigue (\(E_{\text{fatigue}} \to 1.0\)) and financial motives, the simulator forces digital staff agents to generate complex, non-linear anomalies. These behavioral defects serve as a reference baseline for verifying internal investigation and Forensic audit algorithms:
6.1 Consumables Siphoning (Class-2 Resource Fraud)
The simulator replicates workplace misappropriation of expensive, commercial-grade professional cleaning agents. Staff agents execute this behavior by introducing an activity block into the process graph and transmitting a DTO to the write-off endpoint /api/v1/wms/chemistry/write-off for Class-B SKUs under the pretext of “expiration date reached” or “container depressurization/leakage.”
The external Process Mining system captures an anomalous spike in spoilage acts on a specific floor. Concurrently, the Forensic analysis modules correlate these transactions with historical room occupancy logs to uncover a clear discrepancy: the volume of written-off cleaning supplies physically exceeds the maximum allowable consumption baseline per square meter.
6.2 Inventory Lifecycle Tampering (Class-3 and Class-4 Resource Fraud)
The system models a pattern where employees intentionally conceal the loss or theft of property (such as Class-2 bathrobes and linens, or Class-4 remote controls and premium tableware). When an exhausted staff agent discovers a shortage upon a guest’s checkout, instead of initiating an official penalty fee—which requires time-consuming documentation and lowers their personal KPI metrics—the agent transmits a fraudulent DTO record indicating that the items were sent to the laundry facility.
This creates “virtual inventory” balances within the database. Inside the Process Mining transaction logs, this activity manifests as infinite transactional loops (where a token hangs indefinitely in the IN_LAUNDRY status), which are easily flaggable during automated cross-tabular auditing.
7 Conclusion
Implementing the High-Fidelity simulation paradigm successfully concludes this cycle of research, establishing a foundation for a comprehensive, self-regulating stochastic environment modeled after a tourism enterprise facility. Sacrificing instantaneous execution velocity in favor of step-by-step processing with one-second discretization allows for the integration of full-scale multi-resource ERP matrices and a three-dimensional spatial geofence polygon hierarchy.
The concurrent, emulated interaction of two heterogeneous agent groups (Tourists and Staff) across a single structural infrastructure perimeter generates highly realistic competitive contentions for shared assets and available rooms.
The resulting “dirty” digital footprint—enriched with complex corporate fraud configurations, hidden inventory losses, and legitimate coordinate spatial discontinuities—serves as a robust testing framework for stress-testing modern automated forensic fraud detection engines and Process Mining business intelligence suites prior to their deployment in production environments.
8 References and Standards
To ensure alignment with international tourism management practices, the system entities and simulator metrics conform to:
- UN Tourism Methodology: United Nations definitions regarding tourism products and industries.
- ISO 21401:2018 Standard: Environmental and operational sustainability management systems for accommodation establishments.
Footnotes
Role-Based Access Control (Управление доступом на основе ролей) определяет права выполнения операций в системе WMS/ERP на основе должностного профиля сотрудника.↩︎