Fig. 7. Jumping performance of model 2. (A) Model 2 was placed in a normal starting position. The colored arrows represent ground reaction forces (GRFs) as in Fig. 6. In addition, the GRF per unit N m of torque is shown for hip external rotation (yellow) and hip adduction (blue). (B) The path of the center of mass (COM) of the frog during the ground-contact phase for 500 simulation runs in which the magnitudes of hindlimb torques were randomly varied. A large range of take-off angles was produced from a single starting position. The blue path in B—D represents the simulation run in which the actual torques produced by the real frog were used to drive the relaxed degrees of freedom. The red path represents a simulation run in which hip external rotation was increased fourfold compared with that produced by the real frog during a jump. (C) The vertical V_V and horizontal V_H velocities of the COM for the red simulation run matched those of the real frog (black lines) better than the blue run. However, this required an unphysiological level of external rotation torque. (D) The predicted jump distances for the red and blue runs were smaller than those for the real frog. (E) Unlike model 1, the vertical and horizontal velocities for each simulation run were not correlated with one another (i.e. take-off angle varied from trial to trial). This was because individual torque components produced different ratios of vertical to horizontal GRF (see arrows in A). (F) The magnitudes of the hip (HE) and ankle extensor (AE) torques were significantly (P<0.01; r²=0.69 and r²=0.63, respectively) correlated with variations in the peak horizontal velocity among the simulation runs. Only the magnitude of the hip external rotation (HR) torque was significantly (P<0.01, r²=0.59) correlated with variations in the peak vertical velocity.