TL;DR: Embodying gravity sensation in soft slender robots with a minimalist setup.

TL;DR: Combining vision-based tactile imaging with force–torque sensing enables robots to reliably detect subsurface tendon features during physiotherapy palpation, where force signals alone are often ambiguous, while still maintaining safe and controlled contact.

TL;DR: Gecko-inspired dry adhesives enable gentle robotic grasping in extreme cold, fail below ~-60 °C, and can be reliably restored at much lower temperatures using brief local heating and appropriate preload.

TL;DR: A force-informed deep reinforcement learning strategy enables flexible robotic endoscopes to exploit contact with deformable stomach walls for robust, high-precision navigation in dynamic environments, significantly outperforming contact-agnostic policies and generalizing to unseen disturbances.

We present the design of FBG-based wearable tactile sensors capable of transferring tactile data collected by human hands to robotic hands.

TL;DR: A dual stereo vision system with geometry-guided point-cloud relocation enables accurate 3D morphological reconstruction of millimeter-scale soft continuum robots, recovering fine notch-level details despite low-resolution depth sensing.

TL;DR: A model-free deep reinforcement learning controller enables tendon-driven flexible endoscopes to autonomously navigate in both free space and contact-rich environments, achieving over 90% success within clinical accuracy by retraining policies learned in free space for contact scenarios.

TL;DR: A domain-adaptive Sim-to-Real framework combining IoU-guided image blending and style transfer enables accurate and stable segmentation of oropharyngeal organs from synthetic data, significantly improving real-world performance for robotic intubation despite limited real images.

TL;DR: Transferring the navigation strategy that a redundant soft robot learns from what it has seen in the SOFA-based virtual world to the real world.