Identification of Foot Kinematics Parameters

We present a novel and simple procedure for identifying the kinematics parameters of a human foot. This is important, among other uses, in tuning robotics physiotherapy devices. Foot kinematics at the ankle follows that of 2R serial manipulator but, since internal joint angles are not measurable, conventional calibration is not applicable. In our approach, each axis is moved while the other remains fixed: the end-effector traces a circular arc in space, hence identification reduces to fitting the data on a plane and, then, on a circle. We implemented Nonlinear and Linear fitting. The former employs iterative least-squares: for identifiability, small singular values of the Jacobian are zeroed, while observability analysis proves that accuracy increases when the data are distributed uniformly along the arc. The Linear method is based on fitting by Singular Value Decomposition and coordinate transformation of the data points. Our methods take all foot positions into account, thus improving upon previous work using specific foot-points in a single pose. Unlike other works, we avoid tracing many point-markers on the foot using expensive vision machinery. Our methods are robust and accurate even in special cases (perpendicular or parallel axes), whereas Linear is expectedly faster than Nonlinear fitting.