In order to solve the problem of self-care of disabled or semi-disabled people, an autonomous positioning and navigation intelligent device is studied to assist people to realize the needs of sitting, lying, standing, squatting and walking in multiple position changes. For the problem that the generation value of the heuristic function in the traditional A* algorithm of autonomous navigation is smaller than the actual value, resulting in a larger search space and redundant paths, the research uses different node cost functions to select different models to improve the A* algorithm; by establishing the OPEN table and CLOSE table, in which the OPEN table stores the nodes to be investigated and the CLOSE table stores the non-investigable nodes, the path planning is carried out to enhance the directionality , improve the efficiency of the algorithm, the improved A* algorithm implements restricted query paths under the completion of regular expression constraints to achieve restricted, global and optimal autonomous path planning, which improves the practical application of autonomous navigation.