Abstract: Objectives In order to improve the ability of ship path planning and obstacle avoidance in complex Marine environment, and improve the economy and safety of ship navigation, a new method based on improved DDPG algorithm is proposed. Methods The priority experience replay mechanism based on path importance score is introduced to enhance the utilization efficiency of important experience in the learning process. The self-attention mechanism is integrated into the actor-critic network, and the network design is optimized by using the idea of Dueling deep Q network to improve the network's ability to perceive environmental characteristics and enhance the estimation accuracy of value function. Results In the simulation of the East China Sea and the Indian Ocean, compared with DDPG and A* algorithm, the improved algorithm significantly optimizes the path length, the number of inflection points and the number of collisions. For example, in the East China Sea, compared with DDPG algorithm, the improved algorithm reduces the path length by 0.75%, the number of inflection points by 26.92%, and the number of collisions by 15.80%. Compared with the A* algorithm, the improved algorithm reduces the path length by 4.59% and the number of inflection points by 42.42%. Conclusions The improved algorithm is superior to DDPG algorithm and traditional A* algorithm in different complexity Marine environments, which strongly confirms that the improved algorithm has significant advantages and strong universality, and can provide a reference for intelligent decision-making of ship navigation. Key words：Ships；route planning；Obstacle avoidance；A* algorithm；DDPG algorithm