Abstract: The grid method is widely used in mobile robot environment modeling and path planning. In the narrow and long space such as coal mine roadway environment, the size of obstacles, robots and space is out of balance. Environment map, taking twice the longest distance from the robot center point to its own contour as the grid size, is easy to lead to the failure of path planning. In order to solve this problem, this paper proposes a grid map modelling method for a narrow and long space, its main idea is take the size of the obstacle that affects the passage of the robot as the grid size and the robots occupies multiple grids on the map; improves the traditional A^* algorithm by using the obstacle collision detection function; it studies the effectiveness of the improved algorithm in global path planning in long and narrow space by simulation and carried out the path planning experiment in the actual similar environment. The results show that the proposed grid modeling method and the improved A^* algorithm can obtain a collision free route from the starting point to the end point, and can complete the path planning of complex environment in long and narrow space.