Abstract: In order to realize a safer and more efficient water quality inspection operation of an aquatic-air amphibious robot arriving at multiple target points across a pond, and in order to reduce the time as well as the economic cost of the traditional water quality inspection mode, it is very important for people to plan a reasonable path for the robot. The traditional A * algorithm, as a widely used classic algorithm, but it has some shortcomings when facing the complex environment of multi habitat robot operations, so aiming at the traditional A* algorithm path zigzagging, low search efficiency, and unable to consider the different constraint characteristics of multi-amphibious robots, this study proposes a path planning algorithm for water-air amphibious robots that improves the A* algorithm. Firstly, the algorithm needs to collect the distribution of obstacles and the different height information of obstacles by using binocular camera, and use the collected information to create a 2.5-dimensional grid map of multiple water bodies using a computer; secondly, the different speed and energy consumption ratios of the robot were measured through real machine experiments in both flight and ship modes, then the evaluation function of A* algorithm adds different energy consumption, time and safety cost of amphibious robots, and adjusts the different weights between them to obtain the different initial paths accordingly; and then dynamically allocates the weights between energy consumption, time and safety cost to improve the path planning of A* algorithm. Then, the heuristic function of A* algorithm is improved by dynamically allocating the weights of energy consumption, time and safety cost to speed up the search efficiency of A* algorithm, grouping water quality detection target points located in different water bodies and the target cost function is used to prioritize all water quality detection targets, so as to realize the path planning of multiple target points in a partitioned fishpond; then the paper optimizes the initial paths by adding additional airborne mode-switching paths, deleting the redundant paths, and adopting the B-spline algorithm to optimize the initial paths, and finally generates a path that can connect multiple waters. Finally, the initial path is optimized by adding additional airborne mode switching path points, removing redundant path points and using the B-spline algorithm to generate a three-dimensional smooth trajectory that can be connected to multiple waters and contains multiple water quality inspection points. The simulation test results show that compared with the traditional A * algorithm and the land air A * algorithm, iteration times of the improved A* algorithm are reduced by 70.04% and 68.07%, and the path length is reduced by 35.44% and 7.6%, the total turning angle is reduced by 83.63% and 8.65%, and the number of dangerous nodes is reduced by 86.67% and 33.33% compared with the traditional A* algorithm and the land-air A* algorithm. The real waterway experiment shows that the iteration times of the improved A* algorithm are reduced by 84.89% and 83.78% compared with the traditional A* algorithm and the air-land A* algorithm, the path length is reduced by 12% and 0.6%, the total turning angle is reduced by 73.21% and 22.1%, the number of dangerous nodes is reduced by 84.62% and 80% compared with the traditional A* algorithm and the land-air A* algorithm. The improved A* algorithm proposed in this paper can plan a safe and smooth path through multiple target points in multiple pieces of water, which can effectively improve the efficiency of water quality detection in partitioned fishponds in comparison with the traditional approach and the traditional water quality testing vessel, it can also provide a reference for the research on the autonomous navigation of the multi-habitat robot.