Period: 01. 10. 2025 – 30. 09. 2027

Funding SOURCE: MORS

PROJECT PARTNERS: UM, GUARDIARIS, ADD d.o.o., UL FE, IGEA d.o.o, MEDIUS d.o.o.

GEMMA CONTACT: prof. dr. Domen Mongus

GRAPO AI

Abstract:

Development and demonstration of a simulation graphics engine for tactical simulations, where several key functionalities are managed and supported by Artificial Intelligence (AI) and Machine Learning (ML) algorithms.A core component is the generation of synthetic imagery and video of training scenarios based on input text descriptions. This approach can significantly shorten the planning and creation of scenarios, placing these tools directly into the hands of soldiers and instructors. Closely related is the capability to generate graphics and textures for building facades using currently available data (e.g., Google Street View). These facade textures can be further augmented with physical properties, allowing objects to be assigned specific characteristics and vulnerability models (modeling damage and destruction). The system also implements speech analysis of conversations conducted during training between trainees and instructors, specifically in the Slovenian language.The solution should be based on existing open-source AI solutions in the domains of image/video generation and speech recognition. These models will be fine-tuned with domain-specific imagery and video to achieve high-quality results tailored for virtual training in the defense industry.The final implementation must demonstrate the solution on at least one graphics engine currently used by the Slovenian Armed Forces for tactical training purposes (e.g., anti-tank warfare).