Architectural Overview
The XRisis platform architecture integrates CORTEX2 enabling technologies within Mozilla Hubs open-source virtual reality framework, creating microservices-based system supporting multi-user immersive training scenarios across heterogeneous device types. The architecture separates concerns across distinct service layers: communication infrastructure (Rainbow CPaaS), multi-player synchronisation (Unity server and WebSocket coordination), immersive client applications (VR headset and desktop Unity builds), facilitator control interfaces (web-based scenario management), AI dialogue services (CEA CoVA integration), avatar-based video (DFKI VCAA integration), and automatic documentation (Linagora summarisation). This separation enables independent scaling, component replacement without cascading system impacts, and distributed deployment across European cloud infrastructure in Germany and France ensuring data sovereignty compliance.
Mozilla Hubs Foundation
Mozilla Hubs provides foundational WebXR capabilities through microservices architecture comprising: Hubs Client (front-end application built with WebGL and WebRTC enabling real-time communication and 3D rendering), AWS backend managing room creation and user sessions, Spoke (web-based 3D scene editor for environment customisation), and Reticulum (Phoenix Framework networking layer in Erlang facilitating WebRTC peer-to-peer connections). XRisis customised Hubs Client through plugin injection enabling integration with Rainbow SDK, Ready Player Me avatar system, and ConvAI conversational AI capabilities, extending base Hubs functionality whilst maintaining core platform stability. The Reticulum networking layer handles non-voice-video traffic including avatar transformations, authentication, scene updates, and user permission validation through pub-sub system (Phoenix Channels) enabling efficient communication coordination. This architecture leverages Erlang VM strengths for scalability and fault tolerance essential for reliable professional training delivery.
CORTEX2 Service Integration Specifications
Rainbow CPaaS Integration: Alcatel Lucent Enterprise's Rainbow Communication Platform integrates via C# SDK providing native Unity support for secure real-time voice and video. Integration architecture embeds Rainbow communication channels within Unity application scenes, enabling spatial audio positioning when participants communicate in shared virtual environments and conventional stereo audio during one-to-one calls or facilitator briefings. The system handles dynamic communication topology supporting individual work, small group conversations, full team coordination, and facilitator-participant interactions without requiring manual call setup breaking scenario immersion. Rainbow's enterprise infrastructure provides reliability essential for professional contexts where communication failures would destroy scenario realism and compromise learning objectives.
Conversational Virtual Agent Integration: CEA's CoVA platform powers AI dialogue characters appearing as call participants within Rainbow communication channels. Integration employs vector embeddings of Action Contre la Faim documentation (Standard Operating Procedures, Emergency Management materials, training content) grounding AI responses in organisational knowledge rather than relying solely on general language model training. The system supports character personality configuration defining communication styles, emotional ranges, knowledge boundaries, and conversational behaviours appropriate to stakeholder roles within scenario contexts.
Video Call Alternative Appearance Integration: DFKI VCAA system (deployed as web-based emulator using MediaPipe and Ready Player Me pending native integration) enables avatar-based video conferencing. Participants share screens displaying avatar representations or employ virtual camera software injecting avatar video into Rainbow calls as camera feed replacement, maintaining privacy whilst preserving visual presence and non-verbal communication channels.
Summarisation Agent Integration: Linagora's automatic summarisation service processes Rainbow call recordings through speech-to-text transcription and language model summary generation, producing PDF reports capturing key discussion points, decisions, disagreements, and action items. Integration architecture stores recordings temporarily for processing before deletion, maintaining privacy whilst enabling post-exercise analysis and evidence-based debrief conversations.
Infrastructure and Deployment Architecture
The platform deploys on European-owned cloud infrastructure providers in Germany and France, prioritising data sovereignty and GDPR compliance essential for humanitarian sector operations handling sensitive information about organisational capacity, operational plans, and individual performance. A Unity server coordinates multi-player engine synchronisation across all participant endpoints (XR head-mounted displays, desktop applications, web browsers), maintaining consistent game state and managing network replication of user actions, environment changes, and scenario progression. WebSocket server coordinates object distribution between simulation environments, managing inject delivery (emails, news reports, documents), state synchronisation for shared planning tools, and facilitator control signals triggering scenario phase transitions. Web application provides facilitator interfaces for simulation coordination and inject management, serving as front-end for WebSocket server whilst enabling observation of participant activities and session recording for subsequent analysis.
Content Delivery Network, Web Application Firewall, and reverse proxy infrastructure protect platform services from distributed denial-of-service attacks, manage TLS termination, and route traffic appropriately across microservices components. Database services store user profiles, scenario content, learning records, and session metadata with encryption at rest and appropriate access controls limiting data exposure according to role-based permission models.
Scalability and Performance Considerations
The distributed architecture supports horizontal scaling where increased participant load can be addressed through additional server instances rather than requiring vertical scaling that eventually encounters hard limits. Unity server instances can be multiplied to support concurrent sessions without shared state dependencies creating bottlenecks. WebSocket servers similarly scale horizontally with load balancing distributing connections across available instances. Rainbow CPaaS infrastructure scales automatically through Alcatel Lucent Enterprise's managed services without requiring XRisis platform scaling coordination. Database and storage services employ AWS-managed scaling mechanisms handling capacity increases transparently.
Performance optimisation focuses on minimising client-side computational requirements enabling participation from modest hardware: scene complexity restrictions limiting polygon counts and texture resolutions, asset preloading reducing runtime download requirements, level-of-detail implementations adapting visual fidelity to available rendering capability, and graceful degradation disabling advanced features on resource-constrained devices whilst maintaining core functionality access.
Security and Privacy Architecture
Authentication employs secure methods with two-factor email confirmation, federated identity enabling single sign-on across platform components whilst maintaining security boundaries, and role-based access control ensuring users access only capabilities appropriate to their assigned roles. Communication security leverages Rainbow's enterprise infrastructure providing end-to-end encryption for voice and video channels, whilst WebSocket connections employ TLS encryption preventing eavesdropping on scenario data and participant actions. Data protection measures include encryption at rest for stored records, automatic deletion of recordings after summarisation processing, anonymisation of aggregated analytics, and explicit consent workflows for data collection respecting GDPR requirements and organisational privacy policies.
Integration Testing and Validation Architecture
The platform implements comprehensive integration testing validating not just individual component functionality but coordinated operation under realistic loads with diverse participant behaviours. Test scenarios simulate concurrent users, varied network conditions, exploratory interactions triggering edge cases, and failure modes including disconnections, component unavailability, and resource exhaustion. Monitoring infrastructure captures detailed logs enabling issue tracing to specific components, performance metrics identifying bottlenecks and degradation patterns, and user analytics informing interface refinement and feature prioritisation.
Conclusion
The XRisis platform architecture successfully integrates multiple CORTEX2 enabling technologies within coherent system supporting humanitarian emergency management training across diverse deployment contexts. The microservices approach enables flexibility, scalability, and component evolution whilst the Mozilla Hubs foundation provides proven WebXR capabilities and community support. Future development will focus on optimising performance for resource-constrained contexts, enhancing integration depth with CORTEX2 services as they mature, and expanding platform capabilities based on operational validation feedback whilst maintaining architectural principles enabling continued evolution.
Reference Documentation
Complete technical specification available through XRisis project deliverable D1.2 Specification and Implementation Plan, submitted September 2024. Architecture documentation maintained in platform repository with API specifications, integration guides, and deployment instructions.
Related
Partners
Industries
Products