COMBO – Combining Planning, Self-Organization and Reconfiguration in Robot Ensembles for ScORe Missions

移动多机器人系统在科研和工业领域受到越来越多的关注。其中，推动这一发展的原因是，最近通过先进的电子技术、小型化和更强大的机载控制系统，这些设备的控制变得更加容易，而且价格合理。虽然搜索和救援领域的应用多年来在不同环境的众多项目中受益于这一发展，但现在许多其他领域也认识到多机器人应用的潜力。这些包括环境研究，关键基础设施的分布式监控，或处理重大灾难（例如，化学事故，洪水，重大火灾）等。我们确定了与所有这些应用程序相关的常见步骤，并将其称为ScORe任务，涉及搜索、持续观察和反应任务。现有的方法已经表明，空中机器人（uav）和地面车辆的组合适合于应对ScORe任务。然而，这些方法通常侧重于一个ScORe任务的孤立任务，或者针对特定的应用程序进行定制。由于在专用应用程序或环境中对机器人的功能有特定要求，因此这种专业化是必要的。除了任务需求不同外，集成系统在处理ScORe任务时还面临着不确定性。常见的不确定性的例子是机器人在运行时的缺陷，对集成必须工作的初始（环境）设置缺乏清晰度，以及在运行时的开发。这些不确定性使得很难提前计算出整体中每个机器人的完整计划（包括任务调度和分配）。复杂的ScORe任务显然需要在运行时进行调整。我们的项目旨在为机器人整体开发一个参考系统架构，以处理这些任务，重点是空中机器人和移动地面车辆。因此，我们的项目遵循了将经典规划方法与自组织机制相结合的想法，使机器人集成能够适应集成以及环境中不可预见的变化，并做出适当的反应，无需耗费时间的重新规划阶段，并且在ScORe任务的所有任务中尽可能自主。为了使机器人集成具有必要的自由度，我们提出用语义知识增强现有的模块化硬件组件，使其具有自我意识。这允许集成通过在运行时启动硬件和软件级别上的重新配置来推断和优化当前情况。

Mobile multi-robot systems gain increasing attention in research and industry. Among others, this development was driven by the fact that controlling these devices had become much easier recently through advanced electronics, miniaturization and thus more powerful onboard control systems, available at affordable prices.While applications in the domain of Search and Rescue (SAR) have profited from this development for years in a multitude of projects from different environments, the potential of multi-robot applications now is recognized in many other domains. These include environmental research, distributed surveillance of critical infrastructure, or dealing with major catastrophes (e.g., chemical accidents, flood, major fires), among others.We identified common steps, relevant in all of these applications, and call them ScORe missions involving Search, continuously Observe and React tasks. Existing approaches have shown that ensembles of aerial robots (UAVs) and ground vehicles are suitable for coping with ScORe missions. However, those approaches often focus on isolated tasks of one ScORe mission or are tailored to particular applications. This specialization is necessary due to specific requirements to the robots’ capabilities in dedicated applications or environments. Besides different task requirements, the ensemble is faced with uncertainties when dealing with ScORe missions. Examples for common uncertainties are defects of robots at run-time, lack of clarity regarding the initial (environmental) setting the ensemble has to work in, and its development during run-time. These uncertainties make it hard to calculate a complete plan (including task scheduling and allocation) for each robot of the ensemble in advance. Complex ScORe missions clearly call for adaptation at run-time.Our project aims at developing a reference system architecture for robot ensembles to handle such missions with focus on aerial robots and mobile ground vehicles. Therefore our project follows the idea of combining classic planning approaches with self-organization mechanisms to enable the robot ensemble to adapt to unforeseen changes in the ensembles as well as in the environment and to react appropriately, without time-intensive re-planning phases, and as autonomous as possible throughout all tasks of a ScORe mission. To equip a robot ensemble with the necessary degree of freedom, we propose to enhance existing modular hardware components with semantic knowledge, to make them self-aware. This allows the ensemble to reason about and optimize the current situation by initiating reconfigurations on the hardware and the software level at run-time.