"We wanted a sensor that could be added to off‑the‑shelf robots without modifying their internal structure," explains Professor Lu. "Our fiber is thin, flexible, stretchable, and costs less than one US dollar per sensing segment—a fraction of the price of fiber‑optic systems."
The sensor is made by dispersing multi‑walled carbon nanotubes in a thermoplastic polyurethane matrix, then casting the mixture into a thin, 4‑mm‑wide film using solution evaporation. The resulting piezoresistive fiber bandage (PFB) is 250 µm thick, can stretch to 400% of its original length without breaking, and shows a nearly linear resistance‑strain response up to 30% strain. When wrapped helically around the robot's surface, the fiber's resistance changes as the robot bends—the tighter the bend, the greater the resistance shift. Multiple electrodes spaced every 2 cm along the fiber allow the team to sample resistance changes at different points along the robot's body, providing a rich data stream about its deformation.
"The relationship between resistance changes and the robot's 3D shape is complex and nonlinear, especially when you consider the effects of uneven assembly," says Professor Lu. "So instead of trying to derive an analytical model, we used a data‑driven approach." The team divided the robot into three sections, each covered by a separate segment of the fiber bandage, and trained three small fully‑connected neural networks—one per section—to map resistance changes to the arc parameters of a piecewise‑constant‑curvature model. For a single actuated section, the network achieved an average absolute error of just 3.05° for bending angle and 7.49° for the rotation plane angle. For a multi‑section robot following an irregular closed trajectory, the errors were 5.05° and 13.23°, respectively—sufficiently accurate for real‑time intraoperative guidance.
To test the system in realistic scenarios, the researchers inserted the sensor‑equipped robot into three 3D‑printed anatomical phantoms: a sinus model (bending restricted to the distal segment), a bronchus model (bending of distal and middle segments), and a duodenum model (all three segments bent into an "S" shape). In each case, the neural networks successfully reconstructed the robot's shape from the resistance data alone, matching the known configurations of the phantoms. The team then moved to an ex vivo pig intestine, simulating gastrointestinal endoscopy. The sensor‑equipped robot navigated the intestinal canal while the system provided continuous shape feedback, demonstrating its adaptability to real biological tissue.
"This is not about replacing existing sensing technologies—it's about offering a practical, deployable alternative that can be added to robots already in use," emphasizes Professor Lu. "Because the fiber is soft and thin, it adds only 0.6 mm to the robot's diameter and reduces the ultimate bending angle by just 20°, which is negligible for most procedures." The team acknowledges that the current prototype still has too many wires at the base, which can restrict movement, and that the piecewise‑constant‑curvature model can introduce errors in certain regions. Long‑term drift, cyclic durability, sterilization compatibility, and wiring scalability also need further investigation. Future work will focus on streamlining the electrical connections, adopting more precise models to describe the robot's shape, and exploring whether the same sensor can be used for force sensing and force‑feedback control—turning the bandage into a dual‑purpose tool for both shape and haptic perception.
By offering a low‑cost, easily installable, and highly stretchable sensing solution, this carbon‑nanotube bandage could make shape sensing accessible to a much wider range of continuum robots, helping surgeons navigate the invisible interior with greater confidence and precision"
Authors of the paper include Pingyu Xiang, Xiangyu Mi, Hongye Zhang, Fei Wang, Xiong Yang, Yue Wang, Rong Xiong, Song Liu, and Haojian Lu.
This work was supported by the National Key R&D Program of China (2025YFE0113300), National Natural Science Foundation of China (62303407 and T2293720/ T2293724), the State Key Laboratory of Industrial Control Technology (ICT2025A08), and Xiaomi Foundation.
The paper "Enhancing Shape Sensing of Slender Medical Continuum Robot Using Carbon Nanotube Piezoresistive Fiber Bandage" was published in the journal Cyborg and Bionic Systems on Jun. 24, 2026, at DOI: 10.34133/cbsystems.0622.
