EMT/BSSE Programmable Self-Adaptation: A Bio-inspired Approach To Multi-agent Robotic Systems

新兴的计算技术使制造大规模嵌入式多智能体系统成为可能，从庞大的传感器网络到模块化机器人和智能材料。一个关键的挑战是理解如何在单个代理级别对此类系统进行编程，以实现复杂、容错和适应环境的系统级行为。灵感的来源之一是多细胞生物系统(组织、器官和简单生物体)，它们通过大量细胞的分布式合作和感知在不断变化的环境中实现复杂的自我适应。这类系统可以为多智能体系统的设计和编程提供新的生物启发原理。本研究探索了新的计算范例，用于对多智能体机器人系统进行编程，以实现对环境的复杂自适应。这两项主要工作是：(A)开发一种从全局到局部的规划方法，以描述复杂的全球适应目标，并自动推导出可证明是稳健的多智能体控制；(B)开发能够展示自适应结构的、受满意启发的模块化机器人系统。灵感的一个关键来源是细胞用来实现环境响应结构和功能的分散控制策略，例如植物和血管网络中的形状适应，以及简单动物的运动。其目的是利用这些生物启发原理来创造复制生物系统的适应性和响应性的多智能体机器人系统。这项工作具有广泛的应用，从分布式传感器-执行器网络中的鲁棒控制到自适应建筑结构和假肢的开发。这项研究极大地促进了我们对如何设计能够自组织、自修复和响应环境的自主多智能体计算系统的理解。

Emerging computing technologies have made it possible to manufacturelarge-scale embedded multi-agent systems, from vast sensor networks tomodular robots and smart materials. A key challenge is understandinghow to program such systems at the individual agent level in order toachieve system-level behavior that is complex, fault-tolerant, andadapts to the environment. One source of inspiration is multicellularbiological systems (tissues, organs, and simple organisms) thatachieve complex self-adaptation in changing environments through thedistributed cooperation and sensing of vast numbers of cells. Suchsystems can provide novel bio-inspired principles for the design andprogramming of multi-agent systems.This research investigates new computational paradigms for programmingmulti-agent robotic systems to achieve complex self-adaptation inresponse to the environment. The two main thrusts are: (a) thedevelopment of a global-to-local programming methodology fordescribing complex global adaptation goals and automatically derivingprovably robust multi-agent control (b) the development of atissue-inspired modular robotic system that can demonstrateself-adaptive structures. A key source of inspiration is thedecentralized control strategies that cells use to achieveenvironment-responsive structures and functions, e.g. shape adaptationin plants and vascular networks, and locomotion in simple animals. Theaim is to harness these bio-inspired principles to create multi-agentrobotic systems that replicate the adaptability and responsiveness ofliving systems. This work has broad application, from robust controlin distributed sensor-actuator networks to the development ofself-adapting architectural structures and prosthetics. This researchsignificantly advances our understanding of how to design autonomousmulti-agent computing systems that can self-organize, self-repair, andrespond to the environment.