Soft Robotics

Soft robotics is a field that explores robots built from flexible, compliant materials that can bend, stretch, and adapt to their surroundings. Unlike traditional rigid robots, soft robots can move safely around people, handle delicate objects, and navigate tight or cluttered spaces. This makes them particularly exciting for a range of applications: robot assistants in industrial and home environments, inspection and maintenance of hard-to-reach or fragile areas such as the interior of machines and infrastructure, and minimally invasive medical procedures. 

Soft robotics is one of the most interdisciplinary areas, drawing together mechanics, materials science, electronics, sensing, control theory, and machine learning. The core challenge, and the reason the field is so compelling from a research perspective, is that soft robots are far harder to model and control than their rigid counterparts. Therefore, c Current research in the field spans modeling, sensing, actuation, control, learning, and experimental validation, with a strong focus on making these systems more predictable, controllable, and deployable in real-world applications. 

At ETF Robotics, our work in soft robotics connects theory, algorithms, and hardware. We study both soft-bodied continuum robots and articulated soft robots with variable stiffness, with the goal of making them more precise, reliable, and intelligent. A central research direction is the development of control and learning methods that can cope with the uncertainty, nonlinearity, and high dimensionality of soft robotic systems, while still providing guarantees on stability and precision. This work is supported by an open-source soft robotics platform that we have developed to enable extensive experimentation possibilities.  

We combine fundamental and applied research in soft robotics, with emphasis on model-based control, adaptive methods, and stiffness-regulated operation.  

Our research includes several interconnected directions: 

• Modeling of continuum soft robots, by using curvature-based kinematic formulations and computationally efficient parameterizations. 

• Control of soft robots’ position and stiffness, by using adaptive methods which allow precise and reliable control of soft robots despite incomplete or inaccurate knowledge of their physical parameters and unmodeled dynamics. 

• Estimation of hard-to-measure robot’s parameters such as stiffness by using observer-based methods that learn in real time stiffness parameters without requiring dedicated torque or velocity sensors. 

• Design and development of soft robotic systems centered around the existing low-cost and modular platform which allows implementing different soft robot geometries, actuator configurations, and control strategies. The existing platform at ETF Robotics includes the full stack: mechanics, electronics, firmware, and control software. 

• Soft robotics for confined and hard-to-reach environments, is an application area which will be particularly explored in the Soft-Scout project granted to Maja Trumic that is hosted at TU Delft. More information about Soft-Scout project can be found here: https://maja-trumic.github.io/ 

The laboratory is equipped with an open-source and student-friendly soft robotic platform built by Milos Rasic (more information here: github.com/MilosRasic98/SoftRobotETF), which is especially important because it lowers the entry barrier for engineering students who want to move fast from theory to hands-on experimentation. 

An open-source modular soft robotics platform was built as a low-cost and reproducible setup for experimenting with different highly elastic robot shapes and different actuator mechanisms. It provides support for multiple continuum geometries and materials with interchangeable mounting solutions that make it easy to test new morphologies without redesigning the full setup. Fabrication support, including CAD files for the robot body, molds, actuator components, and PCB manufacturing files, makes the platform especially suitable for project-based student work. 

The laboratory also hosts two generations of highly modular robotic modules driven by variable stiffness actuators. Unlike conventional actuators that operate at fixed stiffness, these modules can actively tune both their position and mechanical stiffness in real time, much like the muscles and joints of the human body. This ability to balance precision and compliance as well as compact design makes them particularly well suited for humanoid platforms, where adaptable, human-like limbs are essential for safe and versatile interaction with the environment. 

Open Source Soft Robotics Platform. Open experimental platform for testing highly elastic continuum robots and tendon-driven actuation, including software, electronics, CAD, and mold files for full reproducibility. Link 

M. Trumić, C. Della Santina, K. Jovanović, A.Fagiolini, “Adaptive Control of Soft Robots Based on an Enhanced 3D Augmented Rigid Robot Matching,”IEEE Control Systems Letters, 2021. This work develops a 3D PCC-based modeling and adaptive control framework for soft-bodied robots, emphasizing robust trajectory tracking under significant model uncertainty.

M. Trumić, K. Jovanović, A.Fagiolini, “Decoupled Nonlinear Adaptive Control of Position and Stiffness for Pneumatic Soft Robots,”The International Journal of Robotics Research, 2021. This work presents simultaneous closed-loop control of motion and stiffness for antagonistically actuated pneumatic soft robots and validates the method experimentally.

A. Fagiolini, M.Trumić, K. Jovanović, “An Input Observer-Based Stiffness Estimation Approach for Flexible Robot Joints,” IEEE Robotics and Automation Letters, 2020. This paper addresses online estimation of flexibility torque and stiffness for electrically driven variable stiffness systems, with reduced sensing requirements.