At present, with the development of science and technology, exploration and research of the deep sea in the scientific field has become a hot spot. So the development of human occupied vehicle(HOV) has been accelerated. HOV has played an increasingly important role in marine biology, marine science and marine resource. In order to further meet the urgent needs of scientists in full ocean depth field, the research of full ocean depth HOV has been gradually carried out. However, the demand of underwater high-precision real-time navigation of HOV also brings new challenges. This thesis is based on the subject of full ocean depth HOV control technology research in National key R&D plan for the 13th five-year plan. In this paper, the integrated navigation system and the data fusion algorithm of underwater robot are studied systematically. Then the background and environment of navigation of full ocean depth HOV and the problems of noise and asynchronous fusion are deeply analysed. After the three aspects of integrated navigation, including data, model and method, this thesis makes a careful study on the problems of integrated navigation system of full ocean depth HOV and proposed algorithms to solve the problems. These algorithms that can raise the accuracy of the integrated navigation system and provide accurate navigation information have been certified by simulation experiment. It has important significance for tracking and controlling the track of HOV. The main contents of this thesis are as follow: 1、 Firstly, the integrated navigation system and background of full ocean depth HOV are analysed and the noise problems and asynchronous fusion problems are pointed out. Then the causes of these problems are studied. 2、 Three aspects of integrated navigation system of underwater robot are studied, which are data aspect, model aspect and method aspect of fusion algorithm respectively. The advantage and unadvantage of Kalman filter, extended Kalman filter, unscented Kalman filer and particle filter are analysed, which provide the theoretical basis for proposing integrated navigation algorithms of full ocean depth HOV. 3、A particle filter algorithm based on BOR resampling algorithm was proposed after studying the noise problem of full ocean depth HOV. Then the simulation manifests its effect on noise problems and improving accuracy of navigation. 4、A LR-UKF algorithm based on machine learning and UKF is proposed after studying the asychronouos fusion problem of full ocean depth HOV. And the feasibility of the algorithm is vertified by simulation. Then comparing with the DF-UKF, the accuracy of integrated navigation algorithm is more prominent.