Aiming at the challenge of mobile robots struggling to effectively avoid dynamic obstacles in unknown and complex environments, an obstacle avoidance method that integrates Kalman filtering and an improved Dynamic Window Approach (DWA). First, the Kalman filter algorithm is utilized to predict the motion trends of dynamic obstacles based on real-time position data. The predicted points are then expanded in conjunction with the actual positions of the obstacles, preventing collisions caused by intersections between the robot’s trajectory and the predicted obstacle path. Next, to address the local optimal problem that may arise when the robot and dynamic obstacles move synchronously during avoidance, a new evaluation function based on the angular direction is introduced. This function improves the scoring of the opposite side of the target trajectory, providing more options for avoidance directions. Additionally, to overcome the issue of trajectory oscillation near the target point due to unreasonable weight distribution, an adaptive method is proposed to automatically adjust the weight of the angular direction function based on the angle difference between the robot and dynamic obstacles, thereby adapting to environmental dynamic changes. Finally, the fusion of Kalman filtering with the improved DWA enables the mobile robot to plan its motion trajectory in advance based on the predicted motion trends of obstacles, enhancing the safety of obstacle avoidance. Simulation and experimental results demonstrate that the proposed algorithm can effectively avoid various dynamic obstacles in unknown and complex environments and prevent the robot from becoming trapped in dynamic local minima.