Computer games with the digital characteristics of a non-linear approach have implemented interactive storytelling from the beginning, and today, with high game engine performance, users lead the story with minimal control, and there is an interactive movie genre that allows users to enjoy real-time stories such as movies. Appeared. It requires a lot of manpower and time to produce 3D resources that meet the increased expectations, and a more effective method than the existing production method of predicting, directing, producing, and recording users' needs is needed. Therefore, an advanced sub-unit interaction method was proposed as a way to realize interaction in various situations while reducing the amount of game engine resource production. This is a generation method that actively applies the improved physics engine and artificial intelligence reinforced by deep learning to make the character's movement, dialogue, and natural environment effects in real time in the game engine. As free deep learning libraries such as TensorFlow are activated and GAN learning methods along with CNN and RNN learning methods become generalized, the results of artificial intelligence show a quality that is indistinguishable from the original. In addition, by actively applying the improved physics engine, it is possible to create detailed movements of surrounding objects. The contents considered in the study are technologies that are still implemented, and if they are introduced, 3D resources can be created in real time without directly producing them, and the context setting and dialogue can induce users to naturally follow the flow of the main story. There will be.

This paper analyzes the simulation operation flow of Bullet physics engine. Based on this analysis, four kinds of multi-rigid-body game characters are designed. This paper also profiles the performance metrics such as the CPU utilization, the memory usage, and the computation time by multi-rigid-body character simulations. For the CPU utilization, the Tongs Vehicle is the best and provides 45.1% less than the other character simulations. The computation times of the Four-leg robot and the Dragon are longer than those of the others. The memory usage of the Dragon simulation is the largest, which is average 1.32 times more than the others. Because all parts of Dragon are composed of triangular mesh models in 3DMax. The performance profiling with the criteria such as reducing the computation time and the computing resources, the complexities of the collision shapes, and the number of rigid bodies takes an important role in the design of the multi-rigid-body game characters.

최근 게임이나 시뮬레이터 개발에 있어서 물리기반 시뮬레이션을 통해 사실성을 높인 고품질의 3D 콘텐츠를 개발하고자하는 사례가 많아지고 있다. 본 논문에서는 오픈소스 물리엔진인 ODE를 이용하여 중량물을 취급하는 천장크레인 시뮬레이터의 개발하는 방법을 제안하였다. 그리고 상용 천장크레인 시뮬레이터와의 물리적 특성 비교분석을 통해 오픈소스 물리엔진으로도 사실성 높은 훈련용 시뮬레이터 개발이 충분히 가능하다는 것을 확인하였다. 본 논문에서 제안한 천장크레인 시뮬레이터는 학교 등의 공공기관에서 교육용이나 훈련용으로 충분히 활용할 수 있을 것으로 예상된다.

본 연구는 게임 물리 엔진을 실시간 물리 기술의 관점으로 고찰한다. 실시간 물리 기술이란 물리 시뮬레이션 기술을 게임에 적용하기 위해서 간략화 하는 기술을 말한다. 조사 대상으로 상용 물리 엔진인 Havok Physics SDK와 NVIDIA PhysX SDK를 선택하였고, 오픈 소스기반 물리 엔진인 ODE와 Bullet을 선택하였다. 그 결과 물리 엔진은 강체 역학, 변형체 시뮬레이션, 유체 시뮬레이션을 구현하고 있었고, 실시간 시뮬레이션을 위해서 수식의 간략화, 충돌 처리의 효율성 재고 등 소프트웨어 측면의 기술과 멀티 코어 CPU의 이용, PPU, GPU 활용 등 병렬처리 하드웨어 기술을 사용하고 있었다.