Soft Robotics

The growth of robotics is primarily driven by the exploration of new application areas. Robots are expected to increasingly interact with humans and delicate objects, as well as navigate through changeable, unstructured environments. Conventional robots are typically made of rigid materials and are designed for high forces, precision, and repetitive tasks. As a result, they can pose a significant potential hazard to their surroundings.

Soft robots offer a promising solution to this challenge. Thanks to their flexible materials, they allow for large elastic deformations, limit contact forces, and passively adapt to different shapes and surfaces. This makes them particularly well-suited for applications in human-robot interaction, the medical field, and for grasping delicate or objects of variable shapes, as they might be found during fruit harvesting.

The soft structure, however, presents new challenges for actuation, sensing, modeling, and control. Rigid components and conventional modeling approaches are only applicable to a limited extent. The goal of our research is therefore to improve the proprioception, control, and accuracy of soft robots through integrated sensing as well as modern dynamic modeling and control methods.

Appropriate models are of central importance for the simulation and model-based control of soft robots. In this context, model accuracy, generality, and computational complexity must be balanced against one another. Beam models - such as a piecewise constant curvature (PCC) model or geometrically exact beam theory/ cosserat rods - provide a promising foundation for this purpose:

Conventional actuators can only be integrated into soft robots to a limited extent, as these rigid components conflict with the soft structure. For this reason, alternative actuation concepts are being developed and applied in soft robotics. Typical examples include tendon-based actuation, the use of shape memory alloys (SMA), and fluid-based actuation:

Sensor feedback is essential for precise trajectory tracking, reliable positioning, and sophisticated grasping applications in soft robots. In addition to appropriate actuation, knowledge of the robot’s current position and orientation is particularly important. This requires shape sensors to detect the robot’s curvature and force sensors to control the gripping force. Conventional sensors are only of limited use for this purpose, as they are generally neither designed for large elastic deformations nor capable of reliably tracking them. Furthermore, they are often too rigid, which impairs the compliance and mobility of soft robots.

One sensor concept investigated for detecting segment bending is a resistive curvature sensor based on electrically conductive foam. When a soft robot segment equipped with this sensor bends, the electrical resistance of the foam changes. The measured resistance values can then be analyzed using neural networks to infer the bending angle of the segment.

Other concepts include capacitive force and torque sensors, resistive force sensors, and fiber-optic curvature sensors.