SBIR Phase I: Composing Digital-Twins from Disparate Data Sources

点击翻译按钮获取中文摘要

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project relates to the creation of a "digital twin," a real-world system projected into spatially-computed environment such as virtual and augmented reality. This technology creates the infrastructure necessary for the application of virtual and augmented reality in industrial workplaces, at scale. By making the full-scale roll out of these technologies possible, the technology seeks to impact human health and safety, operational efficiencies, and environmental risk reduction for process operations facilities, such as oil refineries and chemical plants. The long term impacts of the technology may also enable automation and optimization, improving their efficiency, security, and safety. Such facilities are critical infrastructure and play a significant role in the national economy. The availability of this product may also enhance market opportunities for other businesses in the scanning, spatial computing, and training markets. The impact may be further broadened by adapting the process for digital twin production to new domains unrelated to the industrial market.This Small Business Innovation Research (SBIR) Phase I project is advancing knowledge and understanding in both machine learning and spatial computing. This project focuses on a method for digitizing a complex, real-world system, in an efficient manner, sufficient to recreate the captured reality as an interactive digital twin. The primary technical hurdle is the combining of different data sources that describe aspects of a particular real-world system into a single complete description. The initial physical systems being modelled are industrial process operations, but the core methods could apply to other types of systems, including natural systems, such as a rainforest. For industrial process operations, the goal is to encode the entire process operations facilities at the component level, with sub-cm accuracy, at 10% of the current time and cost requirements. To achieve this, this project will combine physical scans with engineering documentation and relational probabilities. Once combined, the model will be used as the basis for a digital twin of the real-world system projected into spatial computed environments, such as virtual and augmented reality. These techniques replace a tedious and impractical static scan and human labor workflow with rapid scans, computer vision, and a combination of procedural and trained algorithms.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.