Warehouse Management System Development Using WPF

A large number of rural distribution centers, manufacturing locations, and remote logistical points have trouble with poor or intermittent internet connectivity to the cloud and have a greatly reduced ability to do real business through the cloud.

This leads to business interruptions, lost data and a complete inability to carry out day-to-day business operations.

And to solve this issue, we developed SmartWare, a desktop application built on Windows Presentation Foundation (WPF) that functions fully offline, synchronizes intelligently when online, and integrates AI-driven insights and IoT sensor data to optimize inventory control.

A 3PL Logistics provider is experiencing many issues:

• A lack of consistent connection to the cloud for syncing files due to weak Internet access.
• Human error while manually entering stock into the system leads to value loss and delays in reconciliation.
• No real time visibility into how a storage bin has been utilized or if there have been forklift movements or not.
• Unfortunately, being forced to share your credentials creates a major security issue that could allow for unauthorized access.

The goal is to create a multi-user Warehouse Management System (WMS) that is resilient, will work with WPF Local database, sync with Azure and be able to support AI/Edge IoT technologies.

Technology Base

• Front End: WPF (modern .NET 9) used for rich data-grid support, integrations to hardware devices (bar-code scanner, RFID readers) and performance of UI on industrial PCs.
• Local Database: SQLite or Microsoft LocalDB (compliance with ACID transaction model while offline).
• Synchronization Engine: Azure SQL Sync Framework or Custom Idempotent API using Hangfire (with retry logic).
• AI Layer: ONNX Runtime in Local System + Python Microservice (Optional).
• IoT Hub: Azure IoT Hub or MQTT Broker via Local Gateway.

Offline-First Principle

Every operation (receive, pick, pack, move) writes to a local PendingSync table. When connectivity is detected, a background SyncService uploads changes in batches with conflict detection (last-write-win or manual override).

4.1 Barcode / RFID Integration

Integration of barcodes and RFID scanner data into WPF projects.

• Use System.IO.Ports to communicate with the USB barcode scanner using either a keyboard wedge type or using a serial type; and use Zebra RFID SDK library to communicate with UHF RFID Readers.
• All real-time scans will lookup against a local SQLite database in less than 50 milliseconds after the scan occurs.
• RFID Portal will automatically register items from incoming pallets on dock doors without line of sight when scanned.

4.2 Stock Tracking & Bin Management

• Warehouse map visualized in WPF DataGrid + Canvas overlay.
• Each bin has QR code; WPF captures bin scans and validates capacity locally.
• Stock levels maintain min/max thresholds – visual alerts (red border) on low stock.

4.3 Cloud Synchronization (Both On-line and Off-line)

• Synchronization strategy: Each online synchronization occurs with a delta every five minutes; image products are binary deltas.
• Conflict resolution procedures for two clients updating an SKU off-line: the manager will review the conflicts in a WFP dialog box.
• All synchronization traffic will occur using TLS 1.3, and all credentials will be stored in the Windows Credential Manager.

4.4 Multi-User Roles (RBAC)

• Hashed passwords and role claims (Administrator, Supervisor, Picker, Receiver) are stored in the local SQLite database.
• Merged views for WPF:
  • Picker -> Only Tasks view of data.
  • Supervisor -> Dashboard with heatmaps.
  • Administrator -> User management + sync controls.
• An audit trail of UserID, Action and Timestamp is stored in the local database for all audits.

5.1 Edge AI – Demand Forecasting

• A capacity-constrained version of an LSTM model was processed locally by the WPF host as an ONNX file.
• The input data consist of local outbound data from the past 90 days.
• The prediction is for the 15 Top selling SKUs scheduled for the upcoming week, with the pre-picked location being highlighted on the WPF to the picker.
• The AI will continue to work without access to the internet, as long as a weekly retraining of the model is performed whenever the cloud has returned to service.

5.2 Anomaly Detection in Cycle Counts

• During inventory counts, a statistical isolation forest model (running in a background Task) flags unexpected discrepancies – e.g., +50 units of an A-item that normally moves slowly.
• WPF shows a warning: "Possible mis-scan or theft – recount bin A-12".

5.3 Voice to Pick (AI Speech)

Optional integration with Azure Speech SDK (offline licenses cached). The picker states, "Bin B3, get 5 units". The WPF NLP extracts this as an intent, checks and updates stock on your local stock then enters an adjustment transaction once it has been confirmed that the appropriate transaction was entered.

6.1 Smart Bin Weight Sensors

• IoT load cells (for heavy bins, e.g., raw chemicals) send weight information to a local gateway through MQTT and also communicate any change in the weight in real-time.
• WPF subscribes to warehouse/weight/sensor_12, if the weight changes more than 2% since the last pick (i.e., no current or previous transaction for that item), then the system will flag the weight change as a potential mis-pick in real time.

6.2 Environmental Monitoring

• Ambient temperature/humidity sensor(s) (e.g., ESP32 + DHT22) connected via an Azure IoT Edge Raspberry Pi.
• If temperature in the frozen goods aisle exceeds -18 degrees Celsius then a critical message is displayed through WPF to the user and no additional items can be picked from the aisle until the supervisor has authorized an exception.

6.3 Forklift Proximity Alerting System

• Ultra wide band (UWB) tags placed on forklifts – with anchors located at either end of racks.
• WPF receives occupancy events – if a forklift enters into pedestrian-only areas during restocking hours; the application logs an incident of safety violation.

Phase 1 (Weeks 1-4) – Offline Core

• Develop WPF MainWindow which includes a DataGrid, TabControl, and scan handlers.
• Create the SQLite database schema for Products, Inventory, Users, and Transactions.
• Implement barcode simulation (Textbox for Scanner Input).

Phase 2 (Weeks 5-6) – Sync Service

• Create SyncService (BackgroundService in .NET).
• Access the local SQLite database using SqliteConnection + Dapper.
• Use a REST API client (Polly – Retry+Exponential Back-off) to sync with Cloud.

Phase 3 (Weeks 7-8) – AI and IoT Integration

• Incorporate ONNX Runtime (Download a Pre-Configured Forecasting AI Model).
• Use MQTTnet library to handle IoT message ingestion in WPF.
• Create a basic Dashboard using LiveCharts2.

Phase 4 (Weeks 9-10) - Security and Multi-User

• Secure Login with SecureString (ID and Password) and set View by Role.
• Encrypt the local SQLite database with SQLCipher for security.

The SmartWare WPF system proves that a rich desktop application remains superior for industrial warehouse environments with poor connectivity.

WPF provides hardware integration, local database resilience and Edge-AI/IoT platform capabilities to deliver reliable real-time data without requiring any links to a cloud service provider.

Offering a best in class, multi user role-based access, smart alerting and sync engine, we can provide an enterprise grade solution at much less than you would pay if you were to go with a cloud only alternative.

Future Roadmap:

Integrating computer vision to identify damaged product during putaway, and LoRaWAN IoT sensors to enable ultra-low power tracking of inventory!