Figure 5 shows the positions of the real robot controlled by the inverse dynamic strategies with a ramp and splines references.
In each case, complex forms are first resolved into a minimum network of simple splines.
The knots for the splines were determined from inspection of the hazard function for males and females combined.
An approach that was considered was to use interpolation splines to form a smoothed edge.
The two probably best-known and most often applied radial basis functions are called multiquadrics and thin-plate splines, respectively.
The control methods were evolved using genetic programming, once a suitable framework had been set up using splines to compute smooth trajectories.
The viscosity function was then calculated from these splines.
They found that higher-dimensional splines offered significant improvements over standard discretization methods (although for dimensions higher than three, their algorithms are still very time-consuming).
