Abstract: A robot may be at the risk of falling from a high place when it works in an unknown complex environment, so the attitude control ability of the robot in the air should be considered to reduce the damage caused by wrong landing attitudes. When a cat drops from a high space, it can always right itself and land safely. Inspired by this biological phenomenon, the optimal falling trajectory of a free-falling cat robot is studied in a time-optimal manner for the first time, through investigating the first stage of safe landing of a cat, namely the attitude adjustment stage. Firstly, a mathematical model of the robot is formulated based on an axisymmetric dual rigid-body model. Owing to the non-integrable angular velocity, the attitude control problem of the falling cat robot is transformed into a nonholonomic motion planning problem. Considering that the time consumption of attitude adjustment is an important factor to determine the adjustment result, a time-optimal function is built with the virtual torque input instead of the real angular velocity input, and a method to solve this function is set up also. Then, a particle swarm optimization algorithm is proposed to obtain the solution of objective function with the least time consumption of attitude adjustment. Finally, using the optimal solution data, the virtual prototype experiment is carried out in the virtual physical environment, and the righting movement in the air for a falling cat robot is realized. The results show that the time consumption of attitude adjustment is effectively reduced for the free-falling robot by the proposed method.