To mitigate the issue of local optima entrapment, ensure the timely avoidance of dynamic obstacles, and enhance navigation efficiency in path planning, this paper proposes a fusion and improvement of Dijkstra and DWA algorithm for navigation, firstly optimizes the search mode of Dijkstra algorithm, uses bidirectional search to obtain the global optimal path, and then improves the evaluation function of DWA algorithm, and introduces a cost function for predicting escape path to replace the traditional safety distance evaluation function. The obstacle avoidance ability of the robot was improved, and finally the algorithm of fusion improvement was analyzed through simulation verification and practical verification. Experimental results reveal that the proposed algorithm significantly improves pathfinding efficiency—reducing iterations by 53.7% and computational time by 11.7% compared to the standard Dijkstra algorithm—while effectively navigating around dynamic pedestrians and obstacles, thereby confirming its overall effectiveness and feasibility.