Abstract: Due to be generated by a unified motion primitive, the mobility of kinedynamic A * algorithm cannot be fully utilized in hybrid aerial-ground robot. So, a kinedynamic A * algorithm was proposed based on the improved motion primitive generation for the hybrid aerial-ground robot. Taking into account the difference in the motion ability of the aerial-ground modes, the performance of the Kinodynamic A * algorithm was arranged to be improved for the hybrid aerial-ground robot based on following three methods. Firstly, the state space acceleration input was determined based on the dynamic boundary to form the diversity land and air modal motion primitives. Secondly, the motion primitives were pruned based on dynamic constraints to ensure them doing not beyond the limits of speed. Finally, the cost function of trajectory optimization was constructed to optimize the smoothness, safety and mobility of the trajectory. The results in the simulation scenario show that the improved algorithm can increase the average speed in the air by 54.1%, reduce the total average path time by 6.7 s, reduce the total path cost by 22.00%, and reduce the algorithm running time by 0.069 s, giving full play to the strong mobility and high endurance of the hybrid aerial-ground robot.