This　study　aims　to　explore　the　driver＇s　takeover　performance　in　different　emergency　takeover　scenes　and　takeover　warnings,　and　to　respond　to　how　to　design　the　emergency　degree　of　takeover　warning.　First,　we　designed　4　takeover　events　based　on　takeover　scenes　(accident　scene　-　emergency,　ramp　scene　-　non-emergency)　and　takeover　request　time　variables　(5s-　emergency,　l0s-　non-emergency).　42　drivers　participated　in　the　driving　simulation　experiment,　and　8　evaluation　indicators　were　obtained　from　four　dimensions:　control　behavior,　eye　tracking,　heart　rate　and　vehicle　performance.　The　results　show　that　the　driver＇s　takeover　performance　is　globally　optimal　in　non-emergency　scenes　and　non-emergency　warnings.　In　non-emergency　situations,　drivers　can　reduce　stress　behavior　and　maintain　good　driving　status.　After　receiving　the　non-emergency　takeover　warning,　the　driver＇s　control　and　psychological　performance　are　better　in　the　emergency　scene.　This　study　can　provide　a　reference　for　a　comprehensive　evaluation　of　takeover　performance　in　different　takeover　scenes　and　warnings,　and　can　also　be　applied　to　the　human-machine　interaction　takeover　warning　design　of　automated　vehicles.　©　2024　IEEE.