This paper addresses the problem of path following in two-dimensional space for underactuated unmanned surface vehicles (USVs), defining a set of guidance laws at the kinematic level. The proposed nonlinear Lyapunov-based control law yields convergence of the path-following error coordinates to zero. Furthermore, the introduction of a virtual controlled degree of freedom for the target to be followed on the path removes singularity behaviors present in other guidance algorithms proposed in the literature. Some heuristic approaches are then proposed to face the problem of speed of advance adaptation based on path curvaturemeasurement and steering action prediction. Finally a set of experimental results of all the proposed guidance laws, carried out with the Charlie USV, demonstrates the feasibility of the proposed approach and the performance improvements, in terms of precision in following the reference path and transient reduction, obtained by introducing speed adaptation heuristics.