A Governance-Aware Private Cloud Architecture for Scalable Multi-Provider Vehicle-Based Multimodal Sensing.

Area of Science:

• Computer Science
• Urban Informatics
• Data Engineering

Background:

• Vehicle-mounted sensing offers high-resolution urban monitoring but faces challenges in data integration, connectivity, privacy, and governance.
• Existing systems often lack robust mechanisms for managing multi-provider data and ensuring privacy-by-design.

Purpose of the Study:

• To introduce a governance-aware private cloud architecture for mobile sensing infrastructures.
• To address challenges of heterogeneous multimodal integration, privacy, and multi-provider governance in urban monitoring.

Main Methods:

• Developed a layered, containerized microservice architecture with asynchronous ingestion and modality-specific processing pipelines.
• Implemented GPU-accelerated object detection and ingestion-time visual abstraction for data minimization.
• Utilized a two-month multi-provider pilot for validation.

Main Results:

• Achieved stable data ingestion without loss and real-time visual inference (~200 ms per frame).
• Demonstrated strict provider-level isolation and up to 95% storage reduction through metadata abstraction.
• Validated the architecture's scalability and privacy-preserving capabilities.

Conclusions:

• The proposed architecture provides a replicable paradigm for scalable, privacy-aware mobile sensing.
• It enables effective multi-actor collaboration for metropolitan-scale smart city deployments.
• The system successfully integrates heterogeneous data while enforcing governance and privacy.