복수 자율주행 클라이언트 동시 성능 평가를 위한 서버 환경 구축

초록

This paper presents an implementation of a platform environment to support the development and performance evaluation of self-driving algorithms by establishing a server infrastructure that can simultaneously perform multiple self-driving agents. As a basic simulation program for this, Grand Theft Auto (GTA) V ? online provided by Rockstar is employed, and its internal data is extracted to generate various sensor output for virtual world simulations. The implemented driving simulator is extended to a server environment by using FiveM where multiple driving and/or navigating agents run concurrently in the same scenario. This server mode driving simulation environment is ideally suited to test and evaluate the performance of self-driving algorithms in a short time by running in parallel.