Software Architectures for Robots Capable of Intelligent Autonomy: A Survey of the State of the Art

Many software architectures have been developed over the past twenty years for autonomous robots. In general, these have focused on the higher-level aspects of the control problem, such as mission planning, navigation, and resource management. Vehicles controlled by these architectures tend to be wheeled robots operating fully autonomously in a quasi-static environment. Despite successes in robot autonomy in such environments, however, complex and dynamical domains are still beyond the capacity of current technology. On the other hand, research in autonomous rotorcraft has mainly focused on problems of stability and control, and not on higher-level issues or mixed-initiative interfaces. It is time for these two lines of research to merge. This article provides an overview of issues connected with architectures for autonomy, a survey of robot control architectures with associated reference list, and some suggestions regarding which seem more appropriate for autonomous Unmanned Aerial Vehicles. None of the existing architectures is ideal for this domain, largely due to their general inability to reason explicitly about dynamics and control.