• Journal of Korea Game Society
• /
• v.8 no.3
• /
• pp.87-95
• /
• 2008

This paper proposes an algorithm for efficient behaviors of boids which freely move and have no predefined position. By finding the kNN and computing the value of behavioral characteristic of a boid approximately, the proposed algorithm improves the conventional spatial partitioning one. To do this, this paper defines and uses the representative boid which has the average direction and position for a group of boids. The proposed algorithm was implemented and compared with the conventional one experimentally. The results of the experimental comparisons show that the proposed algorithm outperforms the conventional one about $-5{\sim}130%$ in terms of the ratio of the number of rendering frames per the second.

• Journal of Korea Game Society
• /
• v.9 no.2
• /
• pp.115-123
• /
• 2009

This paper proposes an algorithm to improve the performance of the spatial partitioning method for flocking behaviors. The core concept is to improve the performance by using the fact that even if a moving entity, boid in flock continuously changes its direction and position, its k-nearest neighbors, kNN to effect on decision of the next direction is not changed frequently. From the previous kNN, the method to check whether new kNN is changed or not is proposed in this paper and then the correctness of the proposed method is proved by two theorems. The proposed algorithm was implemented and its performance was compared with the conventional spatial partitioning method. The results of the comparison show that the proposed algorithm outperforms the conventional one by about 30% with respect to the number of frames per a second.

• Proceedings of the Korean Information Science Society Conference
• /
• 2003.10a
• /
• pp.73-75
• /
• 2003

게임 세계에 존재하는 수많은 보이드들의 지능적인 집단 행동을 모델링하기 위한 방법으로 플로킹 기법이 많이 사용되고 있다. 특히 생태계에 존재하는 객체들의 행동을 재현하기 위한 연구가 활발하게 진행되고 있는 실정이다. 따라서 본 연구에서는 생태계에서 흔히 볼 수 있는 먹고 먹히는 관계를 퍼지 논리를 통해 게임의 보이드 행동을 모델링하고 구현한다.

• Journal of the Korea Society for Simulation
• /
• v.18 no.4
• /
• pp.49-58
• /
• 2009

Top-down approach in the intelligent robot research has focused on the single object intelligence however, it has two weaknesses. One is that has a high cost and a long spending time of sensing, calculating and communications. The other is the difficulty of responding to react changes in the unpredictable environment. we propose the collective intelligence algorithm based on Bottom-up approach for improving these weaknesses and the applied agent model and verify by simulation. The Modified Flocking Algorithm proposed in this research is the algorithm which is modified version of the concept of the Flocking (Craig Reynolds) which is used to model the flocks, herds, and schools in the graphics or games, and simplified the operation of conventional Flocking algorithm to make it easy to apply for the number of group robots. We modeled the Boid agent and verified possibility collectivization of the Modified Flocking Algorithm by simulation. And We validated by the actual multiple mobile robot experiment.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.14 no.2
• /
• pp.195-200
• /
• 2004

Computer games use an intelligent method called flocking for boids' group behavioral modeling. Flocking can naturally model group behavioral patterns of unpredictable forms such as birds and fishes using some computer resource. In this paper, we implemented an ecosystem which is composed of predator and prey for group behavioral modeling of real underwater ecosystem. Also fuzzy logic is applied to implement instinct desire of ecosystem elements. As the result, we confirmed that the model can overcome breakdown of ecosystem and model naturally ecosystem behavior.