Self-Organized Resource Allocation for Reconfigurable Robot Ensembles

Abstract

Mobile robot systems usually are designed, built, and programmed for dedicated use cases. Consequently, especially for unmanned aerial vehicles diverse applications result in very heterogeneously designed robots. To overcome this need for specialization, we propose to dynamically adapt the robots' capabilities at run-time. This is done by connecting and disconnecting hardware modules providing those capabilities, i.e., re-allocating resources within the robot ensemble. Thereby, no longer individualized robots have to be designed for different tasks. Instead, the system is enabled to adapt its hardware configuration to changing requirements. For calculating necessary adaptations, i.e., solving the resource allocation problem, we propose a heuristic, market-based approach that exploits the possibility to decompose the resource allocation problem and distributively finds a solution. We show that our approach outperforms a centralized one especially when increasing the problem size in terms of agents, tasks, and relevant capabilities while providing the same quality.