Constructing Synthetic Environments that satisfy the content, fidelity, functional and performance requirements and expectations of a widening range of applications, users and client systems, in a cost effective and timely manner, challenges todays Synthetic Environment (SE) production capabilities.
Rapidly expanding data availability, increasing processing and rendering performance, a heightened awareness and familiarity with digital, virtual environments in everyday life, sets an expectation and demand that Synthetic Environments will provide high quality, content rich, authentic representations of the real world.
Synthetic Environment production pipelines have evolved to incorporate a range of processes and activities that reflect the progressive convergence of traditional M&S with geospatial, gaming and other relevant domains. But despite employing efficiency enhancing techniques, such as procedural content generation, these production pipelines typically still require significant effort and time to accomplish the task.
This paper will explore how emerging AI, machine learning and deep learning could be applied to a typical SE production pipeline, based on the recently approved SISO Reuse and Interoperation of Environmental Data and Processes (RIEDP) standard.
Beyond M&S, AI is already being applied to the types of task and activity that crossover into SE production – geospatial data processing, construction of gaming assets and environments, computer vision, autonomous vehicles, behavioural modelling. The paper will identify potential AI techniques and technologies that may be applicable to each stage of the production process, consider the feasibility of applying AI across the entire pipeline, from design through to build of the complete SE, and the opportunities AI can offer to SE generation and SE users.