Full Virtual Workflow to support Automation & Autonomy Applications for Commercial Vehicles including AI

For supporting the development of guidance, automation and up to autonomous systems in agriculture, both based on classical image processing methods and on AI, a robust and extendable simulation system is highly beneficial. In this paper we present a general workflow for appending an existing framework based on a common game engine by implementing new assets, virtual environments according to the required scenarios and additional tooling and software support for providing simulation capabilities in the respective application development areas. We execute this process in order to create synthetic data for training of AI for crop row detection, developing assets for the John Deere 612R self-driving sprayer as well as numerous assets of corn crops in multiple stages of growth. We show that, among other benefits for the business, this process is effective in improving performance of a crop row detection algorithm.