Compared　to　open　roads,　highway　tunnels＇　continuous　semi-enclosed　wall　structures　exacerbate　sight　distance　challenges　for　vehicles　equipped　with　automation　systems　(AV)　in　curved　segments.　However,　AV＇s　adaptability　to　existing　road　geometry　in　tunnels,　which　was　primarily　tailored　for　traditional　human-driven　vehicles,　remains　inconclusive.　Therefore,　this　study　aims　to　investigate　the　influence　of　horizontal　curved-tunnel　geometry　on　AV＇s　available　sight　distance　(ASD)　and　analyze　AV＇s　sight　distance　safety　(SDS).　To　this　end,　we　established　a　virtual　co-simulation　platform　for　emulating　AV＇s　ASD　in　diverse　tunnel　scenarios,　which　include　light　detection　and　ranging　(LiDAR)-based　sensing　configurations　and　tunnel　geometry　combinations.　On　this　basis,　the　important　features　related　to　ASD　were　extracted　using　the　recursive　feature　elimination　algorithm　and　several　widely-used　machine-learning　models　were　developed　to　predict　ASD　estimation.　The　Shapley　additive　explanation　analysis　was　conducted　on　the　most　performant　model　to　interpret　the　feature　effects.　Moreover,　reliability　analyses　under　varying　driving　automation　levels,　speeds,　and　pavement　conditions　were　conducted　to　quantify　the　probability　of　noncompliance　with　SDS　in　scenarios　focusing　on　circular　curves,　followed　by　proposing　an　SDS　evaluation　framework.　The　results　show　that:　i)　random　forest　model　outperforms　other　machine-learning　models　in　predicting　ASD　estimation;　ii)　higher-type　tunnel　geometry,　higher-end　sensing　configurations,　driving　on　the　outside　curve　lane,　and　higher　mounting　height　of　LiDAR　achieve　longer　ASD;　iii)　lower-end　sensing　configurations　cause　ASD　to　be　less　sensitive　to　the　tunnel　geometry;　iv)　SDS　deteriorates　in　the　orders　of　moist,　dry,　and　wet　pavement　conditions　and　automation　levels　4,　3,　1,　and　2;　v)　AV　adapt　to　low-type　highway　tunnels＇　horizontal　curves　from　the　sight　distance　perspective,　but　may　fail　in　high-type　designs.　These　findings　shed　light　on　the　impact　mechanism　of　tunnel　curves　on　AV＇s　ASD　and　serve　to　improve　AV＇s　road-oriented　operational　design　domain　and　identify　the　tunnel　segment　that　is　significantly　non-compliant　with　SDS.