<p>As the application of unmanned surface vehicles (USVs) in marine environments increases, existing path planning methods face challenges such as high computational complexity, insufficient real-time decision-making capabilities, and poor adaptability. To address these challenges, this paper adopts the multi-agent soft actor-critic (MASAC) algorithm to investigate cooperative formation path planning for USVs in real-time dynamic environments. Inspired by Dubins path planning methods, a minimum turning radius action space is proposed to address the turning constraints in the navigation process of underactuated USVs, avoiding overly sharp turns, thereby making the path planning more realistic and easier to control. A multi-USV system consisting of a leader USV and follower USVs is constructed, and different reward mechanisms are designed for the leader and follower USVs to enhance the system’s formation capability, with a centralized training and distributed execution framework used to improve sample utilization efficiency. The algorithm enhances strategy exploration through an entropy regularization term. Simulation experiments validate the effectiveness of the proposed algorithm, demonstrating its superiority over multi-agent deep deterministic policy gradient and multi-agent proximal policy optimization algorithms in terms of mission completion, obstacle avoidance, and formation maintenance. Therefore, the MASAC algorithm effectively achieves cooperative formation path planning for multiple USVs in complex dynamic environments, providing a more efficient and reliable solution for a multi-USV system.</p>