To address the inefficiency and path redundancy in the path planning of composite robots in complex and unknown environments, a novel algorithm integrating an improved A* algorithm with the dynamic window approach (DWA) was proposed. Firstly, by introducing a dynamic inertia weight coefficient, the heuristic function of the A* algorithm was adjusted in real time to achieve acceleration in the early stage of the search and global optimal approximation in the later stage. Meanwhile, the time threshold mechanism for accessing Open List nodes was introduced to prevent the algorithm from falling into local optimal solutions. Grid maps with different sizes were constructed, and the improved A* algorithm was verified through simulation using MATLAB software. Secondly, the improved A* algorithm was integrated with DWA to achieve the synergy of dynamic obstacle avoidance and global path following. Finally, the improved A*-DWA fusion algorithm was experimentally verified in both the ROS (robot operating system) simulation environment and the real environment. The experimental results showed that, compared with the traditional A*-DWA fusion algorithm, the use of the improved A*- DWA fusion algorithm could shorten the planned path by 27.57%, reduce the turning points by 40%, shorten the time consumption by 31.03%, and make the speed variation of the composite robot more stable. The improving A*-DWA fusion algorithm not only has the performance of global optimal path planning, but also exhibits dynamic adaptability, enabling the robot to have a high success rate in obstacle avoidance and meeting the requirements of path planning for composite robots in complex and unknown environments.