Cooperative　trajectory　planning　of　aerial-ground　systems　is　a　fundamental　and　challenging　problem,　which　aims　to　leverage　the　aerial　information　to　assist　the　ground　tasks.　Existing　methods　often　suffer　from　suboptimal　trajectories　or　computation　burden.　In　this　paper,　we　address　cooperative　trajectory　planning　of　aerial-ground　systems　in　which　an　unmanned　ground　vehicle　(UGV)　plans　its　local　trajectory　in　real-time　with　the　assistance　of　an　unmanned　aerial　vehicle　(UAV).　Firstly,　the　UAV　generates　guidance　trajectory　using　nonlinear　model　predictive　control　(NMPC),　which　considers　the　obstacle　distribution　density　as　a　factor　reflecting　the　coupling　effect　of　multiple　obstacles　on　the　UGV,　thereby　avoiding　local　minima　problem　and　improving　the　feasibility　of　the　planned　trajectory.　Secondly,　a　null-space-based　behavioral　control　(NSBC)　framework　is　employed　to　merge　the　guidance　trajectory　into　the　UGV＇s　own　planned　one　as　a　task.　Finally,　an　event　triggering　task　supervisor　is　developed　for　the　UGV　to　decide　the　priorities　of　all　tasks,　which　reduces　the　switching　frequency　of　task　priorities　brought　by　traditional　rule-based　task　supervisors.　Both　simulation　and　experiment　results　show　that　the　proposed　approach　has　superior　trajectory　planning　performance　in　terms　of　trajectory　error,　on-line　computation　time　and　the　success　rate　of　task　execution.