【作者】 孟志刚；

【导师】 邹北骥；

【作者基本信息】 中南大学 ， 计算机应用技术， 2011， 博士

【摘要】 自然界中鸟群、鱼群和蚁群聚集成群及觅食等行为有趣而惊人。受这些现象的启发,分别产生了蚁群优化算法、群组动画中的Boid模型以及人工鱼。这些研究目标和内容都属于不同的学科却有着相似的生物学启迪和内在机理,这一点深深吸引并鼓励笔者进行新的探索和研究。另外和以上研究内容有紧密联系和潜在应用的一个学科是游戏人工智能。本文在总结和分析群组动画、蚁群优化算法和游戏人工智能的国内外相关研究工作的基础上,主要开展了蚁群觅食仿真和动画关键技术研究。取得的主要成果如下:(1)提出了基于直接交互的蚁群觅食仿真。在自然界中蚂蚁之间存在直接交互现象,但在以往的蚁群觅食仿真中忽略了这一点。受蚂蚁之间通过反哺进行的直接交互的现象启发,提出了一个和以往蚂蚁通过信息素交换信息完全不同的直接交互机制。将该机制与传统的信息素交互机制融合以后,得到了两种新的蚁群觅食模型。仿真实验结果表明:新的蚁群觅食模型提高了在有障碍物的场景下找到最短路径的能力。对这个新的蚁群觅食模型抽象以后得到了一个新的蚁群优化算法框架。(2)用群组动画研究方法开展了蚁群觅食动画技术研究。提出了多种搜索策略作为蚂蚁和环境之间的交互规则,结合一个有限状态机和底层的运动构成了一个三层蚁群觅食模型。蚁群觅食动画分别在2D/3D环境中运行进行了验证。特别在3D环境下用两种LOD方法进行了绘制提高了渲染效率,对替身图进行光照着色提高了渲染效果。(3)设计了基于理性决策模型的游戏人工智能行为系统用于游戏的开发,也将这个行为系统用在了蚁群觅食仿真中。游戏AI主要针对的是游戏中智能角色的行为和规划的设计。通过分析当前游戏人工智能中行为建模的方法,本文研究出一套适用于任何动态场景以及复杂非线性故事情节的虚拟角色的行为模型,带给用户以实时交互并包含拟人化智能角色的“真实”游戏环境。通过在智能足球对战平台和Starcraft对战平台上的实验,该方法得到了进一步的验证。更多还原

【Abstract】 The flocks and foraging behaviors of birds, fishes and ant colony are interesting and amazing. Inspired from these phenomena, ant colony optimization algorithm, boid model which pioneered the research of crowd simulation and animation, artificial fishes were produced respectively. I am so attracted to these phenomena that different research content and object have similar biologic mechanism. On the basis of analysis and summary of the relevant research home and abroad, my dissertation is about ant colony foraging simulation and animation and my contribution includes:1. Ant colony foraging simulation based on direct interaction between ants is proposed. Direct interaction between ants exists in real world, but is rare seen in ant colony foraging simulations. Inspired from trophallaxies, a way that ants exchange information directly, a direct interaction mechnasim DACF is proposed to be used in ant colony foraging simulation. Then the mechnasim is combined the direct interaction mechanism with the indirect pheromones interaction mechanism. Based on Wilensky’s ant colony foraging model and Panait’s model, this thesis develops two new models. The simulation results of comparison experiments demonstrate that our new model improves the ability of finding the shortest route in environment with obstacles. Abstracting the principle of the new ant colony foraging simulation model can get a new ant colony optimization framework.2. The research of animation of ant colony foraging with crowd simulation methods is firstly carried out. This model consists of three layers, which together model 2D/3D ant behaviors among static food, obstacle and ant nest. Several search strategies are proposed to represent the interaction rules among ants and the enviroment. Ant colony foraging animation is run in 2D/3D environments separately. Then two LOD methods are used to render ant colony, and Phong illumination model is applied to shade the impostor of ant to get better rendering results.3. A behavior system based on rational decision-making model is proposed to be used in design for games AI. The system is also applied in ant colony foraging simulation. After analysis on the current methods on model creating in games AI, a behavior model for virtual actors is developed. This model is suited for any dynamic scene as well as complicated nonlinear storyline. Thereafter, "real" game environment can be presented to users, where there are real-time interactions with users and the personate intelligent roles. Finally, the advantages of this method are verified by the experiments on the platform of intelligent football competing and Starcraft comprting.更多还原