Modelling of a biped kinematics and analysis of the human walk for a future lower-limbs exoskeleton design.

The objective of this thesis is to show the differences between a walking pattern of a humanoid robot and a future exoskeleton. First of all, a deep analysis of autonomous bipeds has been conducted, underlining many differences between them and a proper human walking. The two main differences are: the hypothesis of flat foot during gait-cycle with respect to the floor; and the modelling of the foot as a whole body (not divided in foot and toes). Starting from the second part of the thesis, there was the need of developing a humanoid model. In order to find an approximation of the human body, a biped model was created. It is a 14-dof model, with links’ measures taken from standard human data and revolute joints representing each articulation. This process begins with the creation of a CAD model, in Solidworks, through which it was possible to generate another Simscape model to carry out simulations and results analysis. The first experiment consists in recreating the inverse kinematics of a biped starting from dr. Ali’s algorithm. Through this approach, each joint variable was calculated starting from the feet trajectories in the Cartesian space, the LIPM (Linear Inverted Pendulum Model) and the assumption of the flat foot. On the other hand, in the second experiment, a two-bodies foot was modelled, with an appropriately tuned intermediate passive revolute joint. The joints’ variables were taken from a complete dataset containing angle measurements of each articulation of a patient with no evident pathological condition walking on a treadmill. Thanks to the Simscape platform, it was possible to calculate the direct kinematics and analyse Cartesian variables such as ZMP and CoG with baso-signals reconstruction. In conclusion, this thesis puts in confront the behaviour the system has when the joints’ movement is imposed by inverse kinematic calculation with the one it has when movements are imposed by measurements on the patient. In future it will be possible to create an exoskeleton adding an online analysis of the EMG signals through deep-learning, in which the patient could close the control loop through his/her movements.