Imagine a continuous soft robotic arm bending around a bunch of grapes or broccoli, adjusting its grip in real time as it picks up the object. Unlike traditional rigid robots, which typically aim to avoid contact with the environment as much as possible and stay away from humans for safety reasons, this hand senses microscopic forces, stretches and flexions in ways that more mimic the compliance of a human hand. Its every movement is calculated to avoid excessive force while completing the task efficiently. At the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL) and Laboratory for Information and Decision Systems (LIDS) labs, these seemingly simple activities are the culmination of complex mathematics, careful engineering, and a vision for robots that can safely interact with humans and delicate objects.
Soft robots, with their deformable bodies, promise a future where machines move more seamlessly alongside people, assisting with care, or handling delicate objects in industrial settings. Yet that same flexibility makes them difficult to control. Small turns or turns can generate unexpected forces, increasing the risk of damage or injury. This motivates the need for safe control strategies for soft robots.
“Inspired by advances in safe control and formal methods for rigid robots, our goal is to adapt these ideas to soft robotics – modeling their complex behavior and embracing rather than avoiding contact – to enable high-performance designs (for example, greater payload and precision) without sacrificing safety or embodied intelligence,” says lead senior author and MIT assistant professor Giulio Jardini, a principal investigator in LIDS and the Department of Civil and Environmental Engineering. There are faculty affiliated with the institute. Data, Systems and Society (IDSS). “This view is shared by recent and parallel work from other groups.”
safety first
The team has developed a new framework that blends nonlinear control theory (control systems involving highly complex dynamics) with advanced physical modeling techniques and efficient real-time optimization, which they call “contact-aware security”. At the heart of the approach are higher-order control barrier functions (HOCBFs) and higher-order control Lyapunov functions (HOCLFs). HOCBFs define safe operating limits, ensuring that the robot does not exert unsafe forces. HOCLFs efficiently guide the robot to its work objectives while balancing safety with performance.
“Essentially, we are teaching the robot to know its limits when interacting with the environment while achieving its goals,” says Kiwan Wong, a PhD student in MIT’s Department of Mechanical Engineering, lead author of a new paper describing the framework. “This approach involves some complex derivation of soft robot dynamics, contact models, and control constraints, but the specification of control objectives and safety constraints is quite simple for the practitioner, and the results are very tangible, as you see the robot move smoothly, react to contact, and never create an unsafe situation.”
“Compared to traditional kinematic CBFs – where it is difficult to specify a forward-invariant safe set – the HOCBF framework simplifies obstacle design, and its optimization formulation accounts for system dynamics (e.g., inertia), ensuring that the soft robot stops early enough to avoid unsafe contact forces,” says Wei Xiao, assistant professor and former CSAIL postdoc at Worcester Polytechnic Institute.
“Since soft robots emerged, the field has highlighted their embodied intelligence and greater inherent safety relative to rigid robots due to passive material and structural compliance. Yet their “cognitive” intelligence – especially safety systems – has lagged behind that of rigid serial-link manipulators,” said co-lead author Maximilian Stölzle, a research intern at Disney Research and formerly a PhD student at Delft University of Technology and MIT LIDS. And visiting researcher at CSAIL says. “This work helps bridge that gap by adapting proven algorithms to soft robots and preparing them for safe interaction and soft-continuum mobility.”
The LIDS and CSAIL team tested the system on a series of experiments designed to challenge the robot’s safety and adaptability. In one test, the arm gently applied pressure to an adherent surface, maintaining a precise force without overshooting. In another, it traced the shape of a curved object, adjusting its grip to avoid slipping. In another demonstration, robots manipulated delicate objects alongside a human operator, reacting in real time to unexpected movements or changes. “These experiments show that our framework is able to generalize to diverse tasks and purposes, and that the robot can understand, adapt, and act in complex scenarios while always respecting clearly defined safety boundaries,” Zardini says.
Soft robots with contact-aware security can certainly be a real value-add in high-risk locations. In health care, they can assist in surgery, providing precise manipulation while reducing risk to patients. In industry, they may handle delicate items without constant supervision. In domestic settings, robots can help with work or caregiving tasks, safely interacting with children or the elderly – an important step towards making soft robots reliable partners in real-world environments.
“Soft robots have incredible potential,” says co-lead senior author Daniela Rus, director of CSAIL and professor in the Department of Electrical Engineering and Computer Science. “But ensuring safety and encoding speed functions through relatively simple objectives has always been a central challenge. We wanted to create a system where the robot could remain flexible and responsive, while mathematically guaranteeing that it would not exceed safe force limits.”
Soft robot models, combining discrete simulation and control theory.
Behind the control strategy is a different implementation of something called the Piecewise Coseret-Segment (PCS) dynamics model, which predicts how a soft robot deforms and where forces accumulate. This model allows the system to predict how the robot’s body will respond to actuation and complex interactions with the environment. “The aspect I like most about this work is the mix of integration of new and old tools coming from different areas such as advanced soft robot models, differential simulation, Lyapunov theory, convex optimization and injury-severity-based safety constraints,” says co-author Cosimo Della Centina, an associate professor at Delft University of Technology.
Complementing this is the differential conservative separating axes theorem (DCSAT), which estimates the distance between the soft robot and obstacles in the environment that can be approximated with a series of convex polygons in a different manner. “Previous distance metrics for convex polygons either could not calculate penetration depth – necessary to estimate contact forces – or yielded non-conservative estimates that could compromise safety,” says Wong. “Instead, the DCSAT metric gives strictly conservative and therefore safe estimates, as well as allowing fast and discrete calculations.” Together, PCS and DCSAT give the robot a predictive sense of its environment for more proactive, safe interactions.
Looking ahead, the team plans to extend their methods to three-dimensional soft robots and explore integration with learning-based strategies. By combining contact-aware security with adaptive learning, soft robots can handle even more complex, unpredictable environments.
“That’s what makes our work exciting,” says Roos. “You can see the robot behaving in a human-like, careful way, but behind that grace is a rigid control framework that makes sure it never goes beyond its limits.”
“Because of their body’s compliance and energy-absorbing properties, soft robots are generally safer to interact with than rigid-body robots by design,” says Daniel Bruder, an assistant professor at the University of Michigan who was not involved in the research. “However, as soft robots become faster, stronger, and more capable, this may no longer be enough to ensure safety. This work takes an important step toward ensuring that soft robots can work safely by offering a method of limiting contact forces across their bodies.”
The team’s work was supported, in part, by The Hong Kong Jockey Club Scholarship, the European Union’s Horizon Europe Programme, the Kulturfonds Wetenschapsbeurzen, and the Rouge (1948) and Nancy Allen Chair. Their work was published earlier this month in the Institute of Electrical and Electronics Engineers. Robotics and Automation Paper,
