TL;DR
Explainable AI boosts trust and cuts errors in ship navigation
Measuring tape-inspired robotic gripper tackles fruit and veggie picking
Tiny hopping robot gets a boost from its bounce
Tactile tech takes a backseat to learning sequence in smart robots
RoboBee sticks the landing
Latest News & Research
by Hitoshi Yoshioka, Hirotada Hashimoto in Applied Ocean Research
The Titanic sank 113 years ago after hitting an iceberg—an accident largely blamed on human error. Today, artificial intelligence is being developed to help ships avoid similar disasters. But can a ship's AI explain why it's making a sudden turn?
That's where explainable AI comes in.
Researchers at Osaka Metropolitan University have created an AI system for ships that not only detects collision risks but also explains its decisions using clear, numerical values. This could be a game-changer as global shipping routes grow more crowded and complex.
“The ability to explain why the AI is doing what it’s doing helps build trust with human crews,” said Professor Hirotada Hashimoto, who developed the model alongside graduate student Hitoshi Yoshioka. The technology may also pave the way for truly unmanned, autonomous vessels in the future.
Grasping and Rolling In-plane Manipulation Using Deployable Tape spring Appendages
by Gengzhi He, Curtis Sparks, Nicholas Gravish in Science Advances
Remember playing with a measuring tape as a kid—seeing how far you could stretch it before it bent? That childhood curiosity inspired a team of engineers at UC San Diego to develop a surprisingly effective robotic gripper made from the same material.
Dubbed GRIP-tape (short for Grasping and Rolling In-Plane), the device uses the springy steel of measuring tape to gently grasp and lift delicate items like tomatoes and lemons. Unlike bulky robotic hands that require complex mechanics, GRIP-tape is compact, flexible, and low-cost—perfect for agricultural tasks or working safely around humans.
Each “finger” is built from two rolls of tape and controlled by motors that allow them to extend, retract, and even rotate objects. Because the tape itself is the gripping surface, it can adjust to different shapes and act like a conveyor belt to move items after picking them up.
“Our goal is to explore unconventional designs,” said senior author Nick Gravish. “Measuring tape is ideal—strong, flexible, and surprisingly gentle.”
Future versions of the gripper may include sensors and AI to let it work autonomously in real-world settings like farms or packing lines. For now, it’s a clever reminder that even childhood toys can inspire high-tech solutions.
by Yi-Hsuan Hsiao, Songnan Bai, Zhongtao Guan, Suhan Kim, Zhijian Ren, Pakpong Chirarattananon, Yufeng Chen in Science Advances
In disaster scenarios like earthquakes, small robots that crawl through rubble can be lifesaving—but they often get stuck or slide off unstable terrain. A new insect-sized robot from MIT might offer the perfect solution: it hops.
Weighing less than a paperclip and smaller than a thumb, this agile robot uses a spring-loaded leg to jump over obstacles and slanted surfaces while using 60% less energy than flying robots. It’s powered by a clever combo of a tiny spring and four flapping wings that stabilize and control its movements.
The robot can leap 20 cm—about four times its height—and navigate tricky surfaces like ice, wet glass, and uneven soil. It’s also durable, withstanding repeated landings without damage, and efficient, able to carry up to 10 times its weight.
Because it’s so light and resilient, it could carry small sensors or even hop onto flying drones for coordinated missions. In the future, researchers plan to equip it with batteries and brains so it can operate autonomously outside the lab.
This innovation could transform how we search for survivors in collapsed buildings—or explore tight, dangerous spaces where humans and larger robots can't go.
Curriculum is more influential than haptic feedback when learning object manipulation
by Pegah Ojaghi, Romina Mir, Ali Marjaninejad, Andrew Erwin, Michael Wehner, Francisco J. Valero-Cuevas in Science Advances
How does a robotic hand learn to grasp and rotate an object—like a ball—without the sense of touch? While it’s long been assumed that tactile sensors are essential for this kind of dexterity, new research from USC’s ValeroLab challenges that idea.
In a study published in Science Advances, researchers used a simulated robotic hand to explore a classic question: Is success more about the hand’s physical sensors (nature), or the way it's trained (nurture)?
Surprisingly, they found that the order in which the robot learns—called the training "curriculum"—matters more than touch itself. Even when simulated with little or no tactile input, the robotic hand could still learn complex manipulation tasks if it was trained using the right sequence of rewards.
“Just like living organisms, robotic systems can be shaped more by their experience than by their built-in hardware,” said lead researcher Professor Francisco Valero-Cuevas.
The findings suggest that smart training strategies could allow simpler, less sensor-heavy robots and prosthetics to learn and adapt in real-world environments—offering a new perspective on how machines (and perhaps even humans) learn to interact with the physical world.
Sticking the landing: Insect-inspired strategies for safely landing flapping-wing aerial microrobots
by Nak-seung P. Hyun, Christian M. Chan, Alyssa M. Hernandez, Robert J. Wood in Science Robotics
Harvard’s RoboBee—a flying robot the size of an insect—has already dazzled researchers with its ability to flap, dive, and hover. Now, it's mastered something just as critical: a safe and graceful landing.
Engineers at Harvard’s Microrobotics Lab, led by Professor Robert Wood, have equipped RoboBee with long, jointed legs inspired by crane flies—those gangly insects often mistaken for giant mosquitoes. These legs help absorb the shock of landing, protecting the robot’s delicate internal components.
Landing has always been tricky for RoboBee, which weighs less than a paperclip. Previous versions simply dropped to the ground, hoping to survive the impact. But thanks to a new controller and leg design, the robot now gently slows its descent, even adjusting for unstable air currents near the surface—an effect similar to helicopter downdrafts.
The legs aren’t just for show. They're modeled on real insect anatomy and engineered to mimic the damping properties that allow crane flies to land softly. It's a perfect example of bioinspired engineering, where insights from nature lead to better tech.
While RoboBee is still tethered to external power and control systems, researchers aim to develop a fully autonomous version. That could open doors to real-world uses—from pollination in vertical farms to search-and-rescue missions in disaster zones.
As postdoc Alyssa Hernandez puts it: “Robotic insects like RoboBee not only advance engineering, but also help us test biological theories in motion.”
Subscribe to Paradigm!
Twitter, Telegram, Telegram Chat, LinkedIn, and Reddit.
Main sources
Research articles
Main sources
Research articles