Yi-Hsuan (Nemo) Hsiao

I'm Nemo, a Ph.D. student in the Soft and Micro Robotics Lab at MIT, supervised by Prof. Kevin Chen.

Inspired by natural insects, my research aims to develop novel actuators/mechanisms and advanced control algorithms for robots to perform multi-functional agile maneuvers comparable to their biological counterparts.

The interdisciplinary nature of my research allows me to collaborate with material scientist, biologist, control scientist, swarm expert, machine learning researcher, and roboticist who provide essential insights.

My work has been covered by CNN, Reuters, CBC, BBC Earth, Veritasium, IEEE Spectrum, MIT Technology Review, and other media outlets.

Prior to MIT, I studied at CityU Hong Kong, Georgia Tech, and TU Delft during my undergraduate years. I also did an internship at Apple in Cupertino. more about me

Google Scholar | Linkedin | Twitter

Bio-inspired sub-gram robots

Research projects

Continuous subgram hopper

This insect-scale continuous jumping robot can traverse various terrains, climb inclined surfaces, leap over tall obstacles, or even hop onto another flying robot.

The soft-actuated sub-gram hopper reduces power consumption by 64% and increases the payload by 10 times compared to its flying counterparts.

This research was published in Science Advances.

Acrobatic soft flyer

This new tiny aerial robot, SoftFly, boasts ultra-long endurance, precisely tracks complex trajectories, and executes acrobatic maneuvers—such as double somersaults in just 0.17 seconds.

Compared to other sub-gram flying robots, the new SoftFly offers an order-of-magnitude increase in lifetime and achieves 2–3× greater lateral velocity while maintaining smaller errors.

This research was published in Science Robotics.

MIT News | Veritasium | Reuters | CBC | New Scientist

Damage-resilient soft flappers

Inspired by the resilience of natural bees upon injury, we developed repair techniques that enable our robots' soft artificial muscles, Dielectric Elastomer Actuators (DEAs), to sustain severe damage.

The soft actuator can thus endure piercing by up to 10 needles and a wing loss, but still generate sufficient power to attain hovering flight.

This research was published on the cover of Science Robotics.

MIT News | IEEE Spectrum | MIT YouTube Short

Deep RL on soft flappers

Motivated by the power of unsupervised learning, we developed an approach to efficiently train RL policies to control the SoftFly.

The resulting neural network successfully achieved multiple zero-shot stable hovering, bridging the Sim2Real gap.

This research was selected as the Spotlight Paper at RoboSoft 2025.

Paper

Perching on wet walls with bio-inspired adhesives

Flying is energetically expensive. To stay elevated, natural insects perch on various objects. Learning from this strategy, we developed a new passive mechanism with bio-inspired adhesives to allow the robot to stay on walls and ceilings.

This novel design enables the robot to perch on and take off from surfaces without additional actuators, under both dry and wet conditions.

With the proximity aerodynamic effect, the operation time of the robot is extended by almost a factor of four.

This research was published in Communications Engineering.

Paper | Video | Blog | Behind the Paper

Deep-learned robust-tube MPC

Model Predictive Control (MPC) is the state-of-the-art control technique, yet computationally expensive. We thus trained a 2-hidden-layered neural network to deep learn the MPC.

The neural net can be executed at 2 kHz in real time and perform similarly to MPC for collision recovery.

This research was awarded Outstanding Dynamics and Control Paper Finalist at ICRA 2023.

Paper | Presentation Video

Yaw control of sub-gram flappers

Yaw control is inherently challenging for flapping-wing aerial robots at the insect scale.

We utilized an inclined stroke plane approach to actuate bi-directional yawing motions and performed flights to track a yaw-angle trajectory on the soft-actuated robot.

This research was published at ICRA 2023.

Paper | Presentation Video

Cooperative beam lifting

Natural bees fly in groups to collectively carry out complex missions. Here, we leveraged SLA printing to efficiently fabricate a swarm of 3D-printed insect-scale robots.

These sub-gram artificial drones can fly cooperatively to achieve beam lifting, a challenging task for a single robot.

This research was published in Advanced Intelligent Systems.

Paper | Advanced Science News

Firefly-inspired luminescent robot

Inspired by fireflies in nature, we embedded electroluminescent particles into our soft artificial muscles and performed flights in a dark environment.

The distinct colors emitted by individual units allow us to track robots' positions and attitudes simply with our smartphones.

This research was published in
IEEE Robotics and Automation Letters.

Paper | MIT News | MIT YouTube | Popular Science

Ceiling effect for perching

An interesting aerodynamic phenomenon, ceiling effect, arises when a spinning propeller approaches an overhang. This proximity effect generates more thrust under the same power consumption.

We quantified this aerodynamic effect with quasi-steady model and found that we could potentially reduce the power consumption by a factor of three while perching under a ceiling.

This research was published in IROS 2018 and IEEE/ASME Transactions on Mechatronics.

IROS Paper | T-Mech Paper | Blog

Delfly Nimble Mini

Followed by the renown Delfly Nimble, we developed a scale-down version of this flapping-wing robot, Nimble Mini.

Nimble Mini has a shorter wing span (two third), larger flapping angle, and a newly-designed yawing mechanism.

Based on the controlled free flights we performed, the Nimble Mini can generate much higher yaw rate while maintaining similar roll/pitch agility compare to the original Delfly Nimble.

Page updated

Google Sites

Report abuse