Sampling of alternatives has been shown to be an operational technique for the estimation of route choice models. In order to apply it, the sampling protocol generating paths must be such that the sampling probability is known, in order to correct for the sampling bias. Few path generation algorithms proposed in the literature allow to compute the associated probability. We present a new path generation method that we are currently investigating. It is based on the Metropolis-Hastings algorithm, so that the sampling probability is actually an input to the method, and not an output. Also, it allows for a great deal of flexibility in terms of the exploration of the set of paths.