Automatic Multi-Robot Control Design and Optimization Leveraging Multi-Level Modeling: An Exploration Case Study

Abstract

In this work, we demonstrate the applicability of a recently proposed automatic synthesis approach for behavioral arbitrators based on Probabilistic Finite State Machines (PFSMs) for a multi-robot scenario. More specifically, a behavior-based controller for a multi-robot exploration scenario is automatically synthesized using a predefined set of basic behaviors and conditions. A key feature of the used synthesis approach is the tailored use of two modeling levels of the scenario, microscopic and submicroscopic, to significantly reduce the computational effort. Furthermore, the modeling is extended by a simplified macroscopic model of the scenario to analytically evaluate the best achievable performance given an ideal controller, taking into account real-world constraints such as limited speed and localization. Taking advantage of the interpretability of the synthesized PFSM-based arbitrators, individualistic and collaborative controllers are analyzed separately to provide insights into the theoretical and experimental effects of collaboration for the considered case study. The obtained results show that the PFSM-based synthesis approach is also suitable for multi-robot scenarios, and in particular that the collaborative solution can compete with a manually designed and fine-tuned solution.