MoBAr: a Hierarchical Action-Oriented Autonomous Control Architecture

Autonomous control in robotics hold promising solutions for a broad number of applications. However, autonomous controllers require highly expertise on heterogeneous technologies, such as Artificial Intelligence Planning & Scheduling, behaviour modelling, intelligent execution and the hardware to control. Connecting these technologies entails several challenges to properly synchronize and verify the robot behaviours to deal with real scenarios. In this article, we present an autonomous controller based on high level modelling to easily enable adaptation of the controller to different robotics platforms and application domains. This controller, called MoBAr , allows on-board planning and replanning for goal oriented autonomy. It relies on technologies such as PLEXIL to model the execution behaviours, or the action oriented planning language PDDL for the domain definition and the planning process. Based on these technologies MoBAr enables an easier deployment of the autonomous controller for different robotics platforms. Moreover, MoBAr enables researching in planning systems applied to robotics domains, as it is possible to replace the PDDL planner and/or domain used without much effort. This fact is demonstrated in the experimental section, in which we demonstrate the adaptability and effectiveness of the controller in three different scenarios, i.e., a robotic arm, an office surveillance robot and an exploration rover while exploiting different planning systems.