Gravar-mail: Toward a framework for levels of robot autonomy in human-robot interaction