Depending on the data sometimes transforming the observations, fitting the density, and then transforming back gives appropriate results (and by definition will respect the bounded domain). See "Density Estimation for Statistics and Data Analysis" by B. W. Silverman (pages 29-32).
