In the current context of increasing pressures on marine ecosystems, one of the main challenge of ecology is to be able to conduct accurate and reliable assessment of biodiversity. Until now, studies are mainly performed by divers, which induces high cost and heavy logistic on terrain missions and are limited to few meters depth. An underwater robot could be a solution to most limitations of human-operated observation (divers) in underwater environment, but requires to be specialized according to expert protocols and objectives. This paper presents the design of the control architecture of an underwater hybrid vehicle, where control is distributed among expert (marine biologists) and autonomous system. The analysis of expert protocols drives the control design, resulting in a composition of 'functioning modes', according to the chosen observation strategy and the appropriate control distribution (operator/robot) and actuators allocation (redundant thrusters). We present here the design of the different control laws dedicated to each functioning mode (observation strategy). The performance of the solution is evaluated on the simulator of the ROV Ulysse.