Control ad libitum: a new framework for the design and development of control systems for heterogeneous robot teams

点击翻译按钮获取中文摘要

Teams of mobile robots can efficiently carry out many useful tasks in planetary exploration, emergency response, security, military, and home and office environments. Such a team can be constructed in numerous different ways, and there are many possibilities for the type of robots, control system, communication strategies, and learning algorithms. The goal of this research is to establish an alternative generic framework for the design and development of control systems for heterogeneous mobile robot teams to overcome some fundamental challenges of real-world applications. The new framework introduces three key features to the control system architecture, namely a) dynamically distributed processing through agent persistence and propagation, b) scalable hierarchical control requiring no prior knowledge of team membership, and c) social learning applied to communication and knowledge sharing in a heterogeneous team. The end result will be a team of robots that is highly modular and fault tolerant, flexible in terms of grouping and task coordination, and capable of improving its collective performance over time. In this research, the following series of tasks will be performed by such a team: i) basic navigation, ii) team localization, iii) exploration and mapping, iv) object discovery, and v) object collection. These tasks will be carried out in a cluttered environment that is unknown to the team at the outset of each trial.