The work completed during the realisation of the project started with understanding the problems of modelling and simulation of underwater robots. Understanding the properties of the robotic system and preparing a simulation framework for it are both essential steps to ensure smooth and efficient control design and successful testing on the real platform. Due to the lack of suitable simulation software, that would fulfil the needs required in dynamical simulation of the whole underwater vehicle-manipulator system, being in contact with environment, the researcher decided to develop his own solution. This work became a project of its own, taking into account the complexity of the task. The developed advanced, open-source simulation tool called Stonefish, designed with marine robotics in mind, is one of the most significant achievements of the project. It was released to the public through the GitHub service and licensed under free software General Public License 3.0 (GPL 3.0). The author believes that the software will have a significant impact on the marine robotics society interested in realistic simulation of motion of underwater and surface robots as well as underwater vision. Next, the researcher has investigated available solution to the main problem of the project, by analysing literature about underwater robots as well as industrial robots. As expected, there was little work done on compliant control in underwater manipulation and no experimental results were ever shown in this area. The researcher had to base his search for the suitable algorithm on the well established theory for fixed based manipulators, used in the industry. Thanks to the developed simulation tool it was possible to implement and safely test candidate algorithms. The simulation mimics the real system and communicates with a clone of the software architecture of the real robot, which allows for minimum changes in code when moving it to the real robot. Finally, the developed control algorithms were implemented on the real robot, together with other solutions needed to complete the planned experimental trials, like vision based localisation algorithms, sensor drivers, mission managers, etc. Moreover, for each experimental scenario the researcher had to prepare the experimental rig and the end-effector of the robot. For the most important trial, the non-destructive testing scenario in the CIRS test tank, the researcher had to build a test rig composed of a section of large diameter PVC pipe and an aluminium supporting structure, as well as an end-effector imitating the inspection head, equipped with a force/torque sensor. Similar preparations were needed to repeat the experiments in the OSL tank, during the researchers secondment. The robot successfully performed the desired tasks, producing experimental results never published before and thus extending the state of the art in autonomous underwater manipulation. Moreover, during the project period the researcher took part in preparing two proposals for continuing the research of which one was successfully funded. Apart from the research tasks, the author has taken part in multiple public outreach activities, among them workshops with school students, presentation of the lab facilities and the topic of underwater robotics to high school students, Researcher’s Nights, industrial exhibitions. The researcher has also taken part in international conferences and workshops, where he has given presentations of his work and other works of the CIRS lab.