MIT researchers have developed a groundbreaking control system that allows soft robots to safely interact with people and objects while maintaining precise control of their movements. The innovation, created by teams from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) and Laboratory for Information and Decisions Systems (LIDS), represents a significant leap forward in making soft robots practical for real-world applications where they must work alongside humans.
The Challenge of Controlling Flexible Machines
Soft robots offer tremendous potential advantages over their rigid counterparts. Their pliable, deformable bodies can bend around objects, adjust grip pressure in real-time, and interact more safely with humans. However, this flexibility creates a significant control challenge – the same characteristics that make soft robots adaptable also make them difficult to predict and manage.
Assistant Professor Gioele Zardini, lead senior author and principal investigator at LIDS, explains the core problem: ‘Small bends or twists can produce unpredictable forces, raising the risk of damage or injury.’ This unpredictability has limited soft robots’ practical applications, despite their promising characteristics for healthcare, caregiving, and industrial settings.
The Mathematical Solution: Contact-Aware Safety
The MIT team’s innovation lies in their mathematical approach to robot control. Their framework combines nonlinear control theory with advanced physical modeling and real-time optimization to create what they term ‘contact-aware safety.’ At the system’s core are two mathematical constructs: high-order control barrier functions (HOCBFs) and high-order control Lyapunov functions (HOCLFs).
HOCBFs create defined safety boundaries that prevent the robot from applying excessive force, while HOCLFs guide the robot toward completing its assigned tasks efficiently. Together, these functions allow the robot to balance performance with safety constraints in real-time.
Kiwan Wong, PhD student and lead author, describes it as ‘teaching the robot to know its own limits when interacting with the environment while still achieving its goals.’ Unlike previous approaches, this system accounts for complex dynamics like inertia, allowing the robot to anticipate when it needs to slow down to avoid unsafe contact forces.
Practical Demonstrations: Safety in Action
The team validated their approach through a series of challenging experiments designed to test both safety and adaptability. In one demonstration, a soft robotic arm pressed against a compliant surface with precise force control, never exceeding safe pressure limits. In another test, the arm traced the contours of curved objects, continuously adjusting its grip to maintain contact without slipping or applying excessive force.
Perhaps most impressively, the robot successfully manipulated fragile items alongside a human operator, responding in real-time to unexpected movements or nudges without compromising safety. These experiments demonstrate the system’s ability to generalize across diverse tasks while maintaining strict safety boundaries.
Daniela Rus, CSAIL director and co-lead senior author, emphasizes the significance: ‘You can see the robot behaving in a human-like, careful manner, but behind that grace is a rigorous control framework ensuring it never oversteps its bounds.’
The Technical Underpinnings
Two key technical innovations enable this breakthrough. First, the Piecewise Cosserat-Segment (PCS) dynamics model provides a differentiable implementation that predicts how the soft robot will deform and where forces will accumulate during movement and interaction. This predictive capability allows the system to anticipate the robot’s physical response to commands before executing them.
Second, the Differentiable Conservative Separating Axis Theorem (DCSAT) estimates distances between the soft robot and obstacles in the environment. Unlike previous approaches, DCSAT provides conservative distance estimates that prioritize safety while remaining computationally efficient enough for real-time control.
Cosimo Della Santina, associate professor at Delft University of Technology and co-author, highlights the interdisciplinary nature of the solution: ‘The aspect that I most like about this work is the blend of integration of new and old tools coming from different fields like advanced soft robot models, differentiable simulation, Lyapunov theory, convex optimization, and injury-severity–based safety constraints.’
Real-World Applications and Future Directions
The implications of this research extend across multiple sectors. In healthcare, soft robots with contact-aware safety could assist in delicate surgical procedures, providing precise manipulation while minimizing risk to patients. In manufacturing, they could handle fragile components without constant supervision. In home settings, robots could help with caregiving tasks, safely interacting with elderly individuals or children.
University of Michigan Assistant Professor Daniel Bruder, who wasn’t involved in the research, notes: ‘As soft robots become faster, stronger, and more capable, [inherent safety] may no longer be enough to ensure safety. This work takes a crucial step towards ensuring soft robots can operate safely by offering a method to limit contact forces across their entire bodies.’
Looking ahead, the MIT team plans to extend their methods to three-dimensional soft robots and explore integration with learning-based strategies. By combining their contact-aware safety framework with adaptive learning capabilities, soft robots could potentially handle even more complex, unpredictable environments while maintaining their safety guarantees.
Bridging the Intelligence Gap
This research addresses a critical gap in soft robotics development. While soft robots have long been praised for their ’embodied intelligence’ – the inherent safety and adaptability provided by their physical design – their ‘cognitive intelligence’ has lagged behind that of traditional rigid robots. This new control framework helps close that gap by adapting proven algorithms from rigid robotics to the unique challenges of soft, continuum robots.
The work, published in IEEE’s Robotics and Automation Letters, represents a significant step toward making soft robots reliable partners in real-world environments where they must interact safely with humans and delicate objects. By combining mathematical rigor with practical engineering, the MIT team has created a system that maintains the inherent advantages of soft robotics while addressing their most significant limitations.