Virtual AI training and observation environment based on multiplayer game design

Abstract:
Vulcanos aims at providing an extendable and flexible multiplayer environment for virtualization and AI training purposes. Build on the basis of a sophisticated game engine, it combines features of large multiplayer games with the robotic control software pitasc. Current possible application scenarios range from a visualization of near indefinite large scenes, over physics based testing grounds up to complex AI training possibilities. The latter benefiting from the interconnection of external ROS applications such as pitasc, which can send and receive data directly to and from the running Vulcanos application. In addition, every feature of modern online gaming can be used in the multiplayer environment, allowing the live contribution and cooperation from multiple persons and AI systems from around the world. The results show a stable virtual environment which can be quickly set up, modified and distributed. As the system is based on a well tested game platform, large levels or user counts can be managed without restrictions. In contrast to other similar solutions, the Vulcanos system can quickly be extended in every direction. It sets the foundation for more connected working environments, while keeping the flexibility it needs to support customer-specific adjustments.