Tiling Motion Patches 介绍（一） 翻译

Abstract

Simulating multiple character interaction is challenging because character actions must be carefully coordinated to align their spatial locations and synchronized with each other. We present an algorithm to create a dense crowd of virtual characters interacting with each other. The interaction may involve physical contacts, such as handshaking, hugging, and carrying a heavy object collaboratively. We address the problem by collectingdeformable motion patches, each of which describes an episode of multiple interacting characters, and tiling them spatially and temporally. The tiling of motion patches generates a seamless simulation of virtual characters interacting with each other in a non-trivial manner. Our tiling algorithm uses a combination of stochastic sampling and de-terministic search to address the discrete and continuous aspects of the tiling problem. Our tiling algorithm made it possible to automatically generate highly-complex animation of multiple interacting characters. We achievedthe level of complexity far beyond the current state-of-the-art animation techniques could generate, in terms of thediversity of human behaviors and the spatial/temporal density of interpersonal interactions. 

摘要

模拟多个角色交互具有挑战性，因为角色行动必须 细致协调，使其空间位置和互相保持同步。我们提出一种算法来创建彼此交互的虚拟 密集的人群。相互作用可能涉及物理接触件，如握手，拥抱，和协作运送重物。我们通过收集变形运动的patchs，每一种描述相互作用的多个角色的一节，在空间和时间上平铺他们 来解决问题。拼接的motion patchs 在一种不平常的方式 生成无缝仿真彼此进行交互的虚拟人物。我们分块算法使用随机抽样和确定性搜索组合来解决铺瓷砖问题的离散和连续方面。我们分块算法使能够自动生成高度复杂动画多个交互的角色。我们实现复杂性远远超出当前的最先进的动画技术可以生成的动画，在人类行为的多样性和人际互动的时间空间密度 方面。

Categories and Subject Descriptors(according to ACM CCS):I.3.7 [Computer Graphics]: Three-Dimensional Graphicsand Realism—Animation 

1. Introduction

Simulating multiple characters interacting with each otheris an emerging research topic in computer graphics. Model-ing “interaction” between characters necessitates appropri-ate actions to occur in a carefully coordinated manner in both space and time. Precisely coordinating the spatial location, timing, and action choices of many interacting characters is challenging and often requires prohibitive computation and/or memory resource.

1.介绍

模拟多个角色与彼此交互是计算机图形学中一个新兴的研究课题。建立角色之间的"互动"，就需要采取适当的行动 来应对出现的仔细协调的方式在空间和时间。正是协调许多相互作用字符的空间位置，时间和行动选择是具有挑战性，往往需要高昂的计算和/或内存资源。

We are particularly interested in creating a dense crowd ofvirtual characters that interact with each other. The interac-tions may occur between characters in close vicinity and of-ten involve physical contacts, such as hand shaking, pushing,and carrying a heavy object collaboratively. Each character may perform an arbitrary sequence of actions, which must be spatially aligned and temporally synchronized with other characters. Given a collection of action choices, each character can have exponentially many action sequences and only a small number of action sequences may satisfy stringent spatiotemporal constraints posed by interpersonal interactions.Finding such action sequences simultaneously for many interacting characters is extremely demanding.

我们特别感兴趣创建彼此进行交互的虚拟人物密集的人群。交互作用可能发生在附近的角色之间和往往涉及物理接触件，如握手，推，和协作运送重物。每个角色可以执行的行动，必须空间上对齐和时间地同步与其他角色的任意序列。鉴于行动选项的集合，每个角色可以有呈几何数量级的许多行动序列 并且只有一小部分的的行动序列 可能满足严格的时空约束所构成的人际互动。同时为许多相互作用的角色找到这种行动序列 是要求极高。

We address the problem by collecting episodes of multiple characters. Each episode describes an interesting eventfeaturing a small number of characters for a short periodof time. Tiling episodes seamlessly across space and time would generate a crowd of densely interacting charactersfor an extended period of time. Each episode weaves a col-lection motion fragments to form amotion patch [LCL06].The motion patch has itsentries andexits, which correspondto the beginning and end frames, respectively, of its motionfragments. Two motion patches can be stitched together bycarefully aligning their entries and exits in both space andtime. Our major challenge is tiling patches tightly such thatthe tiling does not have any dangling entries or exits, whichcause sudden appearance and disappearance of characters inthe crowd.

我们解决的问题通过收集多个角色交互的事件。每一个 episode 描述了一个有趣的事件，为一段短时间 设一个小数目的字符。在空间和时间上无缝地平铺episodes将为一个扩展的时间片段 生成一群密集的相互作用的角色。每一个episode编织一个集合motion片段，形成motion patch [LCL06]。motion patch有其的入口和出口，分别对应于的开头和结尾的帧，其运动片段。两个motion patchs可以缝合在一起，通过仔细对准他们的入口和出口在空间和时间。我们的主要挑战是严格的平铺patchs 使得没有任何dangling的进入或退出（使得没有任何突然的进入和退出），在人群中引起的突然出现和消失的角色。

The tiling is a combinatorial problem of layering a collec-tion of motion patches. It also involves continuous optimiza-tion of patch locations for packing patches densely whileavoiding collisions. Feasible solutions are rare and exhaus-tive search is not feasible. Our tiling algorithm uses a com-bination of stochastic sampling and deterministic search toaddress the discrete and continuous aspects of the prob-lem. Our motion patches are deformable to add flexibilityin stitching patches and thus alleviate the difficulty of tiling.The deformable patch allows its motion fragments to warpsmoothly while maintaining the coherence of interpersonalinteraction captured in the patch.

平铺是一个 将motion patches的集合分层 的组合问题。它还涉及到为了packing patches densely（高度压缩patches）不断优化 patch的位置 的同时避免碰撞。可能的的解决办法罕见和穷举搜索是不可行。我们分块算法使用随机抽样和确定性搜索的组合来解决该问题的离散和连续的方面。我们运动的补丁就是变形，以增加在缝合修补程序的灵活性，从而减轻平铺的困难。可变形patches允许其运动片段平滑的变形，同时保持一致性的patch中捕获的人际交往。

Our tiling algorithm made it possible to automatically gen-erate highly-complex animation of multiple interacting char-acters. We achieved the level of complexity far beyond thecurrent state-of-the-art animation techniques could generate,in terms of the diversity of human behaviors and the spa-tial/temporal density of interpersonal interactions. The spa-tiotemporal nature of our tiling algorithm allows us to dealwith both static and dynamic virtual environments. We canalso control the generation of animation by specifying whenand which actions to occur at any specific location of theenvironment. 

我们分块算法使能够自动生成高度复杂动画的交互的多个字符。我们实现复杂性远远超出当前的最先进的动画技术可以生成，人类行为的多样性和人际互动的时间空间密度的水平。我们分块算法的空间时间性质允许我们处理静态和动态的虚拟环境。我们还可以通过指定何时控制的动画生成哪些操作发生在任何特定地点的环境。