Abstract
The presented work belongs to the field of AUV (Autonomous Underwater Vehicle) control software architecture. The control of an AUV is based on a set of embedded computer resources and a set of sensors/actuators which can change according to the robot mission type (change of payload for instance). The robot should be comparable to a general-purpose vehicle adaptable to different tasks, and consequently evolving in accordance with the technological progress or the appearance of new challenging scientific applications.<br /><br />Moreover, the need for autonomy in an environment in constant evolution and frequently unknown requires on behalf of the vehicle to be able, at every moment, to evaluate its state and the state of its environment, in order to make coherent decisions while executing the mission. The realization of such type of autonomous vehicle requires a methodology of design of its embedded software/hardware architecture. <br /><br />Here, we present the software architecture developed at LIRMM for the AUV TAIPAN. This architecture is built respecting several criterions like modularity, evolutionarity and reusability.<br /><br />The architecture is made up of independent components called "modules". Modules have a set of ports that permits data/control flows to be dynamically established between them. It is then possible to program the modules separately, to modify them and to test them before assembling them.<br /><br />This proposed mixed architecture is organised on a hierarchical basis according to two levels: a decisional level containing a Global Supervisor and Locals Supervisors (one for each mode: autonomous, teleoperation, cooperation) and an Executive Level based on a scheduler and a set of modules.<br />A vocabulary based on three types of terms: objectives, sub-objectives, and modules is used within the architecture to express the intention (objectives), the system capacity (modules) and the way to achieve the intention thanks the robot capacity (sub-objectives).<br /><br />The higher level, the global supervisor, is in charge of strategic mission management. Decisions related to the robotic tasks (objectives) to launch at a precise time and to mission planning are taken within this level. It decomposes the mission received from the operator into a sequence of objectives sent to the local supervisor. A local supervisor checks resources availability, reacts to events which require fast response (e.g. obstacle avoiding). It divides each objective in sub-objectives before sending them to the executive level. This executive level works in a periodical way. Several modules within this level are coordinated to configure sensors, to compute controls, to manage instrumentation conflicts. To ensure the respect of low level constraints (e.g. acoustical interfering sensors) a scheduler is in charge of low level modules activity management.<br /><br />To illustrate main aspects of our approach, an application example is developed and tested on the AUV Taipan.