BIC: Biologically-Inspired Robust Space/Time Programming of Sensor/Actuator Ensembles

传感器网络、点对点无线、智能材料、分布式机器人和生物计算等不同领域都设想，网络系统将用数千到数万亿个仅在本地通信的设备填充空间。我们构建如此庞大的、空间嵌入的网状网络的能力正在迅速超越我们对它们的控制能力。生物系统为解决这一问题提供了灵感来源。生物系统表现出退化：它们有多种实现目标的方式。这导致了强大的稳健性和灵活性：这样的系统能够存活并修复损坏，并且只需要对参数进行很小的改变，就可以用制造人手的相同过程来制造狮子的爪子或蝙蝠的翅膀。这个项目的目标是学习和形式化工程原理，这些原理将使我们能够设计和构建具有与生物系统相同的精度、灵活性和健壮性的系统。研究人员打算通过基于几何抽象的编程语言的设计和构建来捕捉工程原理。几何抽象用于隐藏系统由离散个体组成的事实，并允许根据总体所占用空间的行为而不是个体的行为来规范需求。用新语言和工具表达的程序可能具有不熟悉的属性。开发的语言将逐步从模拟器和机器人到生物学。这项研究的一个最终目标是设计出适合和有效的编程生物细胞的语言。生物细胞丰富，有自己的能量生产系统。为了在编程单元中既强大又实用，语言结构必须映射到单元机制，程序必须由可重用的组件构建，并且组件的运行时成本必须是可测量的。模拟器和机器人将逐渐成为目标生物学要求越来越高的试验场。这项工作的计算模型是无定形计算机：系统的组件密集地分布在普通空间中，其中计算和本地通信都很便宜。作为初步实验，原型语言正在开发中。Proto用四个设计原则来表达行为——空间过程、聚合状态、主动过程维护和动态平衡——同时没有明确提及单个设备或它们之间的通信。本研究结果将在工程科学和应用领域产生广泛的影响。无定形计算的灵感之一来自于分子生物学和微制造领域最近的惊人发展，它们有望制造出廉价的微型设备，这些设备可以混合到大量生产的材料中，比如油漆、凝胶和混凝土。这项工作试图揭示的工程原理将使在这种智能材料中编程复杂行为成为可能。另一个应用来自生物计算，这些原理可以应用于程序生物膜，使其具有预定的行为。最后，在调查过程中所学到的原则可能会反映在更好地理解启发他们的生物系统，以及促进生物学家和计算机科学家之间更好的合作。

Diverse fields such as sensor networks, peer-to-peer wireless, smart materials, distributed robotics, and biological computation all envision network systems that fill space with thousands to trillions of devices that communicate only locally. Our ability to construct such large, spatially-embedded mesh networks is rapidly overtaking our power to control them. Biological systems provide a source of inspiration in tackling this problem.Biological systems exhibit degeneracy: they have multiple ways of achieving goals. This results in great robustness and flexibility: such systems survive and repair damage, and with only a small change of parameters the same process that builds a human hand can also make a lion's paw or a bat's wing. The goal of this project is to learn and formalize engineering principles that will enable us to design and build systems that have the same level of precision, flexibility, and robustness as biological systems.The investigators intend to capture engineering principles through the design and construction of programming languages based on geometric abstractions. Geometric abstractions are used to hide the fact that the system is made up of discrete individuals and allow specification of requirements in terms of the behavior of the space occupied by the aggregate rather than the behavior of individuals. Programs expressed with their new languages and tools may have unfamiliar properties.The languages developed will be targetted progressively from simulators and robots to biology. One ultimate goal of this research is to devise languages appropriate and effective for programming biological cells. Biological cells are plentiful and have their own energy production systems. In order to be both powerful and practical in programming cells, linguistic constructs must map onto cellular mechanisms, programs must be built out of reusable components, and the runtime costs of components must be measurable. Simulators and robots will serve as a progressively more demanding testing ground towards targetting biology.The computational model underlying this work is Amorphous Computers: systems with components densely distributed through ordinary space, in which computation and local communication are cheap. The Proto Language is under development as a preliminary experiment. Proto expresses behavior in terms of four principles of design-spatial processes, aggregate state, active process maintenance, and homeostasis-while making no explicit reference to individual devices or to the communication among them.The results of this research will have a broad impact in both engineering science and applications. One inspiration for amorphous computing is the recent astonishing developments in molecular biology and in microfabrication, promising tiny cheap devices which might be mixed into materials that are produced in bulk, such as paints, gels, and concrete. The engineering principles which this work seeks to uncover will make it possible to program complex behavior in such smart materials. Another application comes from biological computation, where these principles could be applied to program biofilms to have predetermined behaviors. Finally, the principles learned during investigation may reflect back into better understanding of the biological systems which inspire them, as well as promoting better collaboration between biologists and computer scientists.