Abstract: Objectives Aiming at the problem that it is difficult to effectively communicate the avoidance intention between autonomous ships and traditional ships in mixed environment, an interactive collision avoidance decision-making method based on Stackelberg game and Chain of Thought is proposed. Methods Firstly, the scenario of ship collision avoidance in mixed environment is described. The autonomous ship and the traditional ship are modeled as a leader-follower Stackelberg game model. The reward functions are designed from the perspective of navigation practice. Secondly, considering the interaction process between ships, a chain of thought-Game collision avoidance decision algorithm ( COT-GCA ) is designed, which includes four sub-modules : state perception, intention sharing, strategy negotiation and collision avoidance decision. Finally, the three-ship and four-ship two-group encounter situations were used for experimental verification. Results The results show that the ships in the two groups of experiments can efficiently understand the avoidance intention of other ships and avoid collision safely, and the response, steering range and resumption of collision avoidance behavior reflect the earlyness, sharpness and stability. The average value of the output efficiency evaluation before and after the decision-making of the decision-making unit evaluation method is 1 and 0.993 respectively, which is close to the optimal, and can explain the high efficiency of the game model in solving the ship 's interactive collision avoidance. Conclusions The proposed model and algorithm can effectively improve the interactive avoidance decision-making ability of ships in mixed environment, and provide theoretical significance for future practical applications.