We　present　a　frontier-driven　autonomous　robotic　exploration　method　on　a　continuous　representation　of　environment.　The　approach　utilizes　radial　basis　function　neural　network　to　build　continuous　occupancy　grid　map.　Parametric　frontiers　are　calculated　directly　by　gradient　field　of　occupancy　probability　distribution,　which　clear　show　division　between　free　and　unexplored　space.　Besides,　the　resulting　frontiers　provide　a　measure　of　quality　automatically.　Simulation　is　present　to　show　the　performance　of　the　proposed　technique.