The multispectral advanced volumetric real-time imaging compositor for real-time distributed scene generation
Joseph W. Morris, Gary H. Ballard, Dennis H. Bunfield, Thomas E. Peddycoart, Darian E. Trimble · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2011
AMRDEC has developed the Multi-spectral Advanced Volumetric Real-time Imaging Compositor (MAVRIC) prototype for distributed real-time hardware-in-the-loop (HWIL) scene generation. MAVRIC is a dynamic object-based energy conserved scene compositor that can seamlessly convolve distributed scene elements into temporally aligned physicsbased scenes for enhancing existing AMRDEC scene generation codes. The volumetric compositing process accepts input independent of depth order. This real-time compositor framework is built around AMRDEC's ContinuumCore API which provides the common messaging interface leveraging the Neutral Messaging Language (NML) for local, shared memory, reflective memory, network, and remote direct memory access (RDMA) communications and the Joint Signature Image Generator (JSIG) that provides energy conserved scene component interface at each render node. This structure allows for a highly scalable real-time environment capable of rendering individual objects at high fidelity while being considerate of real-time hardware-in-the-loop concerns, such as latency. As such, this system can be scaled to handle highly complex detailed scenes such as urban environments. This architecture provides the basis for common scene generation as it provides disparate scene elements to be calculated by various phenomenology codes and integrated seamlessly into a unified composited environment. This advanced capability is the gateway to higher fidelity scene generation such as ray-tracing. The high speed interconnects using PCI Express and InfiniBand were examined to support distributed scene generation whereby the scene graph, associated phenomenology, and the scene elements can be dynamically distributed across multiple high performance computing assets to maximize system performance.