The increasing proliferation of source data poses a significant challenge to the capabilities of Synthetic Environment (SE) generation pipelines to transform this data with the necessary agility, velocity, productivity and affordability required to meet future demands.

Previous research has demonstrated the potential to apply Artificial Intelligence (AI) throughout the entire generation pipeline.  But the ability to capture large volumes of imagery, video, point cloud and other real-world data within days, hours or in real-time places specific demands on the capacity of SE pipelines to extract and process features, attributes and properties from this data, and update or create SE’s at a comparable tempo.

Faced with a similar and potentially overwhelming data glut, fields such as security, autonomous vehicle and Intelligence, Surveillance and Reconnaissance (ISR) systems are performing automated, real-time analysis and computer vision tasks, from facial recognition and human pose estimation to object detection, classification and segmentation.  These are made possible by the application of AI, particularly Deep Learning (DL), to perform critical data processing and analysis in real-time, often on edge devices.

If similar techniques can be applied to SE generation it offers a path to genuinely rapid SE construction.  Today’s largely offline data preparation and processing tasks could be encapsulated into a pipeline that can intelligently analyse, process and exploit input data, extract features, derive and generate content on-the-fly, in an end to end, data to run-time SE construction process.

This paper will examine the state and application of AI to rapid or real-time data processing and analysis. It will assess if, how and where such implementations of AI and DL could be applied to introduce similarly fast, dynamic data exploitation into the SE pipeline, and consider potential issues such solutions may pose in areas such as SE interoperability.