DREAM - data driven ReActive modelling

在一个虚拟和数字物体并存的世界里，重要的是，数字模型的外观和行为都应该能够模仿真实物体，对周围环境做出真实的反应，或者按照我们的愿望进行编程。在这个项目中，我们将开发一个能够实现这样一个目标的建模范例。任何虚拟环境的核心都是代表真实物体的几何模型。在这个研究项目中，第一步是开发一个建模范式来产生一个反应模型；它不仅仅是场景中物体的视觉表现，而且它的行为考虑到它的行为结构，以及它所存在的世界。模型不仅应该能够以有机和物理现实的方式反应并积极参与其他虚拟元素，而且还应该能够回答自我查询。就像智能手机知道自己的位置、状态并完全连接到互联网一样，我们的模型也可以通过元信息进行工具化，从而进一步丰富它们与增强环境的交互能力。例如，人脑元模型可以配备短程传感器，能够在虚拟脑活检过程中准确识别手术工具的位置。我们的方法是建立在隐式建模的概念之上，在隐式建模中，一般拓扑和几何模型可以表示为标量场的等面。通过设计合适的函数，场及其梯度的组合可以产生复杂的反应性模型。在保留表示通用几何形状的能力的同时，这些体积模型允许高效和可扩展的跟踪。

In a world where both virtual and digital objects reside, it is important that not just the appearance but the behaviour of a digital model should be able to imitate real objects, react realistically to its surroundings or be programmable to behave as we desire. In this project we will develop a modelling paradigm capable of delivering such a goal. At the core of any virtual environment are the geometric models that represent real objects. In this research project, the first step is to develop a modelling paradigm to produce a reactive model; not a mere visual representation of the object in the scene, but one whose behaviour takes account of its behavioral structure, as well as the world in which it exists. Not only should a model be able to react and actively participate with other virtual elements in an organic and physically realistic fashion but it should also be able to answer self-queries. In the same way a smart-phone is aware of its position, status and is fully connected to the internet, our models could be instrumented with meta-information that would further enrich their interaction capability with the augmented environment. For example, a human brain meta-model could be equipped with short range sensors capable of exactly identifying surgery tools positions during virtual brain biopsy. Our approach is to build upon concepts from implicit modelling, in which models of generic topology and geometry can be represented as iso-surfaces of scalar fields. By designing suitable functions the combination of fields and their gradients can yield complex, reactive models. While retaining the power to represent generic geometry, these volumetric models allow for efficient and scalable tracking.