Researchers at the Laboratory for Computer Vision of South Ural State University (SUSU), jointly with the Electromechanical, Electronic and Electrochemical Systems Regional Youth Laboratory, have developed an in-house module for simulations of autonomous vehicles. The project at the Youth Laboratory is led by Nadezhda Kuleva, Candidate of Sciences (Engineering), Associate Professor, and Head of the Electromechanical Systems and Electronic Control Units track.
The development is designed as an import-substituting solution intended to replace foreign analogues such as NVIDIA Drive Sim. According to the project team, the module offers a unique level of flexibility, enabling the simulation of highly specific scenarios that are not available in commercial products.
“The vehicle simulator aggregation module, built on the Unity engine with Python scripts for video data processing and control, is primarily being developed for electric vehicles. The system is expected to be tested on autonomous electric vehicles,” explained project author Vladimir Surin, Candidate of Sciences (Engineering), Associate Professor, and Head of the Laboratory for Computer Vision.
In contrast with “closed” foreign systems, the SUSU-developed module allows engineers to freely define any operating conditions and parameters. This capability is critically important for testing electric vehicles, special-purpose machinery, and for adapting algorithms to Russian operating conditions.
Initially, the module is being developed for electric and special-purpose vehicles, including an electric bus that a potential partner of the Laboratory for Computer Vision, the “Heart of the Urals” Advanced Engineering School, plans to acquire as part of an intra-university collaboration. In the near future, however, the modular architecture will make it possible to adapt the system for simulating unmanned aerial vehicles (drones) and ground-based robotic platforms.
The system combines the Unity 3D engine (used to create a realistic virtual environment) with Python-based computer vision and control algorithms. This approach enables real-time testing of complex engineering tasks, such as road surface recognition under challenging weather conditions. All processes are modelled entirely within the virtual environment and do not depend on actual weather conditions or the need to reconfigure a physical testing site.
The development roadmap also includes the integration of large language models (LLMs), including their use for generating unpredictable events and scenarios in a virtual traffic environment, as well as the creation of multi-agent simulations in which multiple autonomous objects interact with one another.
The implementation strategy involves two key stages: comprehensive validation of the algorithms in a virtual environment, followed by testing in a real-world test environment on the SUSU campus, which is currently being digitized.
“The next stage is the testing site,” Vladimir Surin noted. “Once we have completed the core tasks in the virtual space, we will move on to on-site testing. The testing site will be located at the SUSU backyard. Specialists from the Advanced Engineering School are already working on its digital modelling. We are also adapting the system for drones, but this requires more advanced modelling: taking into account air resistance, airflow dynamics, more complex kinematics, and so on. It is a more challenging task, but it is part of our plan.”
At the initial stage, the product will be used for the university’s internal academic research and development. In the medium term, the finished solution is expected to be offered to training centres for specialists’ education. Commercialization for industrial partners will be considered after the main development phase is completed.
The project is being carried out as part of the SUSU’s comprehensive work on intelligent transportation systems and is implemented in close cooperation with the Laboratory for Computer Vision of the “Heart of the Urals” Advanced Engineering School, creating a synergistic effect across the university’s innovation ecosystem.
The project team has been granted a State Registration Certificate for Computer Software No. 2025691910 for the “Vehicle Simulator Aggregation Module on Unity with Python Scripts for Video Data Processing and Control,” officially confirming the relevance and demand for the development.