Abstract In this paper, an innovative hybrid control architecture for real-time control of autonomous robotic vehicles is described as well as its implementation on a commercially available autonomous underwater vehicle (AUV). This architecture has two major components, a
behavior-based intelligent autonomous controller and an interface to a classical dynamic controller that is responsible for real-time dynamic control of the vehicle given the decisions of the intelligent controller over the decision state space (e.g., vehicle course, speed, and depth). The
driving force behind the development of this architecture was a desire to make autonomy software development for underwater vehicles independent from the dynamic control specifics of any given vehicle. The resulting software portability allows significant code reuse and frees autonomy software
developers from being tied to a particular vehicle manufacturer’s autonomy software and support as long as the vehicle supports the required interface between the intelligent controller and the dynamic controller. This paper will describe in detail the components of the backseat driver
architecture as implemented on the Iver2 underwater vehicle, provide several examples of its use, and discuss the future direction of the architecture.