The paper proposes a path planning algorithm based upon Particle Swarm Optimization (PSO) for obstacle-avoidance problem of USV (Unmanned Surface Vehicle). At present, the PSO based path planning algorithm divides the line between start point and goal point into equal parts, and the result path is polyline, which do not conform to the motion features of USV. Besides, many smooth path planning algorithms have been proposed. The Hermite interpolation, cubic splines interpolation and Bezier curve have a tedious process, however the resulting path is not that excellent and smooth. To obtain the optimal smooth path, this paper proposes a new particle model, the arc is the fundamental component of path, and the radius and arc angle are values of dimensions, Comparing with other PSO algorithm and smooth path planning algorithms, the simulation results show the advantage of the Arc Path Planning algorithm in length and smoothness.