Cecilia Laschi: TO CONTROL, OR NOT TO CONTROL- THAT IS THE QUESTION FOR SOFT ROBOTS WITH EI

Описание: Abstract: Embodied intelligence and soft robotics are deeply intertwined, as the deformability of soft body structures play a crucial role in generating emergent behaviours through interaction with the environment. One of the key contributions of embodied intelligence to robotics consists of simplifying control, particularly by distributing the control burden to the body and its interactions with external forces. In this approach, motor behaviours arise naturally from how the robot interacts with its surroundings, reducing the need for explicit, centralized control. However, while soft robots can exhibit emergent behaviours, we still wish to exert more direct control—particularly when it comes to soft robotic arms. For tasks like grasping, positioning the end effector or shaping the robot arm becomes essential. These control tasks present complex challenges that have captivated a growing research community in soft robotics. Unlike traditional, rigid robots, for which consolidated modelling and control techniques exist, soft robots require different modelling approaches, due to their deformable and flexible nature. In response to these challenges, model-free control approaches—often leveraging machine learning—have been widely adopted in soft robotics. These techniques allow robots to learn their own kinematics and dynamics, as well as their interactions with the environment, rather than relying on precise, pre-defined models. While these approaches are effective, recent advances in modelling soft body deformations have led to progress in model-based control. These methods aim to account for the complexities of soft materials and enable more predictable and controlled behaviours. Despite these advancements, one of the most intriguing possibilities in soft robotics is the development of robots that require no control at all. By exploiting the intrinsic properties of soft materials, such as instabilities, some soft robots can achieve movement through purely mechanical means—no electronics or complex control systems needed. These are extreme examples, where a constant input generates movements that are inherently determined by the robot design, demonstrating that control, in some cases, can be entirely outsourced to the robot physical structure. In soft robotics, there exists a wide spectrum of possibilities—from fully controlled systems to entirely control-free and electronics-free ones. Finding the optimal balance between these extremes is one of the key challenges in soft robot design. The trade-off between control levels is the central dilemma in the design process. The challenge is to determine when and how much control is necessary, given the robot tasks and the environment it interacts with.