This thesis addresses a new paradigm for robotic and prosthetic foot-ankle systems. The aim is to study, analyse, and realise artificial feet that replicate key functionalities of the human foot in order to achieve stable and robust locomotion in unstructured environments. The primary guideline follows the principles of soft robotics and underactuation, exploiting intrinsic mechanical compliance to enable adaptive interaction with the terrain, as exemplified by the recently introduced SoftFoot. The thesis begins with the investigation of the deformation of the human multi-articular foot and the dynamics of lower-limb joints across various terrains and obstacles in healthy subjects. The collected kinematic and dynamic data, and their interdependence, establish a foundation for the design of artificial foot–ankle systems. A general model for a tendon-driven adaptive foot is then introduced. The model, comprising two arches, a tendon-like sole, and a toe segment, predicts a two-phase relationship between ankle torque and the angle between arches: an initial compliant phase that allows the foot to conform to surface irregularities and absorb impact, followed by a stiff phase that provides efficient body-weight support. This behaviour is expressed in terms of the geometric and constructive parameters of the foot, enabling application-specific customization. The thesis then addresses the primary limitation of the SoftFoot as a prosthesis, namely the lack of ankle power during push-off. Several approaches are explored while preserving the foot’s adaptability. A simulation environment using a physics engine is first introduced to create a digital twin of the SoftFoot and evaluate novel designs with dynamically simulated walking subjects. The first proposed solution is an ankle–foot prosthesis that extends the SoftFoot architecture with an agonist–antagonist spring set, which recycles gait energy to enhance push-off, resulting in the Agonist–Antagonist SoftFoot Pro. Technical implementations of this architecture are presented, along with their metabolic and kinematic benefits, across various terrains and activities. As the system remains fully passive, it cannot generate more power than what is stored during gait. To enable active power injection, the SoftFoot is integrated with the Open-Source Leg (OSL), a modular robotic knee–ankle system. The combined platform is tested on a transfemoral subject using biomimetic control strategies across even and uneven terrains, and compared against the original rigid feet of the leg. Similar evaluations are performed on transtibial subjects using a task-agnostic controller, followed by trials in a real outdoor scenario. Finally, the agonist–antagonist scheme is extended by replacing the springs with series elastic actuators, resulting in a variable stiffness actuated ankle joint capable of adapting to different activities and terrains while preserving the inherent adaptability of the original SoftFoot. The control of the Variable Stiffness SoftFoot is considered.

From Human Foot–Ankle Dynamics to Bionic Locomotion: The SoftFoot Paradigm

CROTTI, MATTEO
2026

Abstract

This thesis addresses a new paradigm for robotic and prosthetic foot-ankle systems. The aim is to study, analyse, and realise artificial feet that replicate key functionalities of the human foot in order to achieve stable and robust locomotion in unstructured environments. The primary guideline follows the principles of soft robotics and underactuation, exploiting intrinsic mechanical compliance to enable adaptive interaction with the terrain, as exemplified by the recently introduced SoftFoot. The thesis begins with the investigation of the deformation of the human multi-articular foot and the dynamics of lower-limb joints across various terrains and obstacles in healthy subjects. The collected kinematic and dynamic data, and their interdependence, establish a foundation for the design of artificial foot–ankle systems. A general model for a tendon-driven adaptive foot is then introduced. The model, comprising two arches, a tendon-like sole, and a toe segment, predicts a two-phase relationship between ankle torque and the angle between arches: an initial compliant phase that allows the foot to conform to surface irregularities and absorb impact, followed by a stiff phase that provides efficient body-weight support. This behaviour is expressed in terms of the geometric and constructive parameters of the foot, enabling application-specific customization. The thesis then addresses the primary limitation of the SoftFoot as a prosthesis, namely the lack of ankle power during push-off. Several approaches are explored while preserving the foot’s adaptability. A simulation environment using a physics engine is first introduced to create a digital twin of the SoftFoot and evaluate novel designs with dynamically simulated walking subjects. The first proposed solution is an ankle–foot prosthesis that extends the SoftFoot architecture with an agonist–antagonist spring set, which recycles gait energy to enhance push-off, resulting in the Agonist–Antagonist SoftFoot Pro. Technical implementations of this architecture are presented, along with their metabolic and kinematic benefits, across various terrains and activities. As the system remains fully passive, it cannot generate more power than what is stored during gait. To enable active power injection, the SoftFoot is integrated with the Open-Source Leg (OSL), a modular robotic knee–ankle system. The combined platform is tested on a transfemoral subject using biomimetic control strategies across even and uneven terrains, and compared against the original rigid feet of the leg. Similar evaluations are performed on transtibial subjects using a task-agnostic controller, followed by trials in a real outdoor scenario. Finally, the agonist–antagonist scheme is extended by replacing the springs with series elastic actuators, resulting in a variable stiffness actuated ankle joint capable of adapting to different activities and terrains while preserving the inherent adaptability of the original SoftFoot. The control of the Variable Stiffness SoftFoot is considered.
17-nov-2026
Inglese
adaptive foot
legged humanoid
prosthesis
robotics
Catalano, Manuel Giuseppe
Bicchi, Antonio
Grioli, Giorgio
File in questo prodotto:
File Dimensione Formato  
Thesis_Matteo_Crotti.pdf

embargo fino al 21/11/2029

Licenza: Creative Commons
Dimensione 86.78 MB
Formato Adobe PDF
86.78 MB Adobe PDF

I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.14242/385828
Il codice NBN di questa tesi è URN:NBN:IT:UNIPI-385828