This paper is focused on the sampling process for path planners based on probabilistic roadmaps. The paper first analyzes three sampling sources: the random sequence and two deterministic sequences, Halton and sd(k), and compares them in terms of dispersion, computational efficiency (including the finding of nearest neighbors), and sampling probabilities. Then, based on this analysis and on the recognized success of the Gaussian sampling strategy, the paper proposes a new efficient sampling strategy based on deterministic sampling that also samples more densely near the C-obstacles. The proposal is evaluated and compared with the original Gaussian strategy in both 2D and 3D configuration spaces, giving promising results.