CPS: Synergy: Collaborative Research: High-Level Perception and Control for Autonomous Reconfigurable Modular Robots

该项目的目标是开发理论、硬件和计算基础设施，使用户定义的高级任务（如危险环境检查和物体检索）能够自动转换为模块化机器人的可证明正确的控制。模块化机器人由简单的单个模块组成；虽然单个模块的功能有限，但将不同配置的多个模块连接起来，可以让系统执行复杂的动作，如爬坡、操纵物体、在非结构化环境中行驶和自我重新配置（分解成多个独立的机器人并重新组装成更大的结构）。该项目包括(i)通过教育活动和新算法手动定义和填充一个大型感知和驱动构建模块库，（ii）创建自动化工具，为与库组件性能相关的概率指标赋值，（iii）开发一种语法和自动化工具，用于控制合成，对库的不同组件进行排序，以完成更高级别的任务，如果可能的话。或者在任务无法完成时向用户提供反馈；（iv）设计和构建一种新型模块化机器人平台，能够快速、稳健地进行自我重构。这项研究将产生几个结果。首先，它将为制造任何人都能轻松控制的模块化机器人奠定基础。这将为机器人行业带来具有独特功能的新型机器人。其次，这项研究将创造新的算法，将感知、控制和硬件能力紧密结合起来。最后，该项目将创建一个开源基础设施，允许公众为图书馆贡献基本控制器，从而促进机器人和工程的一般研究和社会兴趣。

The goal of the project is the development of the theory, hardware and computational infrastructure that will enable automatically transforming user-defined, high-level tasks such as inspection of hazardous environments and object retrieval, into provably-correct control for modular robots. Modular robots are composed of simple individual modules; while a single module has limited capabilities, connecting multiple modules in different configurations allows the system to perform complex actions such as climbing, manipulating objects, traveling in unstructured environments and self-reconfiguring (breaking into multiple independent robots and reassembling into larger structures). The project includes (i) defining and populating a large library of perception and actuation building blocks both manually through educational activities and automatically through novel algorithms, (ii) creating automated tools to assign values to probabilistic metrics associated with the performance of library components, (iii) developing a grammar and automated tools for control synthesis that sequence different components of the library to accomplish higher level tasks, if possible, or provide feedback to the user if the task cannot be accomplished and (iv) designing and building a novel modular robot platform capable of rapid and robust self-reconfiguration. This research will have several outcomes. First, it will lay the foundations for making modular robots easily controlled by anyone. This will enrich the robotic industry with new types of robots with unique capabilities. Second, the research will create novel algorithms that tightly combine perception, control and hardware capabilities. Finally, this project will create an open-source infrastructure that will allow the public to contribute basic controllers to the library thus promoting general research and social interest in robotics and engineering.