Abstract: [Objectives] Aiming at the difficulties of accuracy maintenance and track control of unmanned surface vessel operating in narrow lakes and culverts, an intelligent predictive control method for trajectory tracking was proposed based on a self-developed torpedo type unmanned surface vessel. [Methods] First, construct the self-developed nonlinear state space model of the underactuated unmanned vehicle; then the intelligent predictive controller is designed, which is based on the design idea of model predictive control and combines with the improved particle swarm optimization algorithm to make online decisions, optimize the performance indicators at every moment and correct the predicted state. Finally, simulation and lake test are carried out to test the tracking performance of the system for reference trajectory, and the tracking performance is compared with that of the linear model predictive controller. [Results] The results show that the designed intelligent predictive controller has fast response speed, small overshoot and good anti-interference. [Conclu-sions] The proposed method can not only be applied to the tracking system of small torpedo unmanned craft, but also can be used as reference for other unmanned surface vessels tracking systems.