An automated negotiation model for human-machine collaborative collision avoidance in intelligent ships

To address the challenge of human-machine collaborative collision avoidance decision-making for intelligent ships in complex and dynamic scenarios, this paper constructs a novel decision fusion framework based on human-machine automated negotiation. Here, the human side is represented by a simulated human agent based on a human-like behavioural model. This framework formalises collision avoidance decision-making as a multi-round, interactive process for reaching consensus. By systematically integrating four core modules – utility evaluation, acceptance judgment, concession generation, and opponent modelling – it facilitates closed-loop negotiation of collision avoidance strategies between the human and the machine. A machine concession mechanism based on fuzzy reasoning is designed to interpret the strategic characteristics of the human opponent and predict its adjustment trends. Simultaneously, a human-like behaviour model is established by introducing the Weber-Fechner law to simulate human decision responses under risk perception, thereby generating adaptive concession rates for both parties. Furthermore, strategic coupling constraints are embedded within the negotiation process to accelerate consensus reaching and prevent strategic oscillations. Simulation results demonstrate that the proposed model can achieve consensus within the stipulated time frame and in fewer negotiation episodes. The generated joint collision avoidance strategy maintains navigational safety while reducing path loss, suggesting improved economic efficiency and stability. This work provides a theoretical reference and a computable solution for human-machine hybrid intelligent collaborative decision-making in complex situations.