Multipotent Systems: Combining Planning, Self-Organization, and Reconfiguration in Modular Robot Ensembles

Mobile multirobot systems play an increasing role in many disciplines. Their capabilities can be used, e.g., to transport workpieces in industrial applications or to support operational forces in search and rescue scenarios, among many others. Depending on the respective application, the hardware design and accompanying software of mobile robots are of various forms, especially for integrating different sensors and actuators. Concerning this design, robots of one system compared to each other can be classified to exclusively be either homogeneous or heterogeneous, both resulting in different system properties. While homogeneously configured systems are known to be robust against failures through redundancy but are highly specialized for specific use cases, heterogeneously designed systems can be used for a broad range of applications but suffer from their specialization, i.e., they can only hardly compensate for the failure of one specialist. Up to now, there has been no known approach aiming to unify the benefits of both these types of system. In this paper, we present our approach to filling this gap by introducing a reference architecture for mobile robots that defines the interplay of all necessary technologies for achieving this goal. We introduce the class of robot systems implementing this architecture as multipotent systems that bring together the benefits of both system classes, enabling homogeneously designed robots to become heterogeneous specialists at runtime. When many of these robots work together, we call the structure of this cooperation an ensemble. To achieve multipotent ensembles, we also integrate reconfigurable and self-descriptive hardware (i.e., sensors and actuators) in this architecture, which can be freely combined to change the capabilities of robots at runtime. Because typically a high degree of autonomy in such systems is a prerequisite for their practical usage, we also present the integration of necessary mechanisms and algorithms for achieving the systems’ multipotency. We already achieved the first results with robots implementing our approach of multipotent systems in real-world experiments as well as in a simulation environment, which we present in this paper.