Mobile AI Edge Computing

---

AI-Powered Insight. Directly Onboard.

Process data, run software containers, and bring edge intelligence directly to the source—without any cloud dependency.

Modern rail and vehicle systems generate huge amounts of data: video streams, sensor telemetry, diagnostic data, operating signals, and more. In many cases, however, it is not practical to transfer this data to the cloud due to bandwidth limitations, latency, or security policies.

Mobile AI Edge Computing Mobile AI Edge Computing solves this problem by moving the intelligence directly into the vehicle. With Parametric's MEC4000 platform, operators can analyze data, roll out AI models, and run customer-specific applications directly on site – reducing latency, ensuring data sovereignty, and increasing system autonomy.

It's not just about computing power. It's about making better decisions faster – right where they're needed.

• For predictive maintenance
  Use onboard analytics to detect anomalies before they turn into failures—without any real-time cloud connection.
• For edge AI video analytics
  Process camera streams directly in the vehicle for object detection, passenger counting, security alarms, or platform monitoring.
• For digital rail infrastructure
  Evaluate track-side sensor signals and system signals locally to minimize response times and reduce network load.
• For operators
  Deploy containerized apps directly to vehicles and manage them centrally—securely, efficiently, and flexibly.

• AI at the edge
  Run machine learning models and inference engines directly on board with an integrated neural processing unit (NPU).
• Custom logic in containers
  Deploy custom software in Docker/OCI containers—from diagnostic scripts to advanced event processing.
• Secure system architecture
  Segment and protect subsystems using VLANs, firewalls, and access policies—all centrally managed.
• Data efficiency
  Transfer only what is truly relevant—pre-filtered, summarized, or event-driven data—and save bandwidth and costs.
• Flexible integration
  Works with cloud, on-premise, or completely offline architectures. You decide where and how data flows.

1. MEC4000 connects to vehicle systems via Ethernet, serial interfaces, or wireless connections. Subsystems are logically segmented and isolated using VLANs and policy-based controls. Sensor and system data is captured and processed in real time by the onboard CPU and NPU.
2. Custom applications run in secure containers – controlled via remote update and orchestration tools.
3. Processed results can be logged locally or transmitted to cloud or control center systems via secure outgoing connections.
4. All remote access is fully traceable, role-based, and can be activated or revoked as needed.

ABB integrates traction converters into trains that must be continuously monitored to ensure performance and safety. However, railway operators — who are not necessarily the system integrators — may not be authorized to access full control or diagnostic data.

In addition to strict access control requirements, the system needed to handle mobile network challenges such as high latency, intermittent connectivity, and the complexity of managing distributed, multi-vendor devices onboard.

Parametric delivered an onboard edge solution based on the MEC4000 Mobile Edge Computer. ABB uses the platform to extract, combine, and evaluate log data from its traction system’s various subsystems. These operations run inside a secure software container.

In addition to log analysis, the MEC4000 synchronizes all attached devices using a common time base and distributes GPS coordinates throughout the onboard network — enabling a unified view of data across the system.

The containerized software environment ensures modularity and easy updates while maintaining full operational independence. All processing takes place onboard; only summarized results and logs are sent to central systems.

• Unified System Monitoring: Distributed subsystems are time-synchronized and location-aware, enabling coherent data evaluation.
• Reduced Data Transmission: Only processed summaries and alerts are sent to the cloud, minimizing bandwidth needs.
• Security & Modularity: Containerization, VPN tunneling, VLAN segmentation, and strong authentication protect data and simplify updates.
• Operational Autonomy: ABB retains full control over software and data without dependency on cloud infrastructure or third-party platforms.

Built for intelligence. Ready for anything.

The solution supports various operating modes:

• Pure edge operation
  All data is processed and stored locally. Ideal for isolated environments or those with limited connectivity.
• Cloud integration
  Transfer selected data for long-term analysis, dashboards, or model training—via secure, exclusively outbound connections.
• Hybrid deployments
  Combine local analysis with higher-level remote monitoring via secure APIs and container orchestration.

Every connection is encrypted. Every application is isolated. Every access is traceable.

Compact. AI-ready. Industry-ready.

At the heart of the solution is Parametric's MEC4000—a powerful edge computer with:

• Basic hardware:
  Onboard LTE, optional 5G/Wi-Fi 6, GPS, managed Ethernet switch
• AI acceleration:
  Integrated neural processing unit (NPU) for ML inference
• Container platform:
  Runs software in Docker/OCI containers with full orchestration support
• Secure architecture:
  VLANs, firewalls, audit logs, and role-based access
• Expansion options:
  2 × PCIe and 2 × USB 3.1 for peripherals or storage
• Certifications:
  EN50155, ECE R10 Rev4 – proven for rail and road