Neuro-Embodied Adaptive Robotics
NEAR Lab
We build robots whose bodies are part of their intelligence.
We explore how physical embodiment, neural computation, control, and environmental interaction can work together to create robots that are more adaptive, efficient, and capable in the physical world.
News
- NEW! Sep 2026 Our paper "Spiking Neural Network Control of a Flapping-Wing Robot on Resource-Constrained Hardware" is accepted at CoRL 2026. Kudos to Rim and Chenrui!
- Jun 2026 Our paper "Koopman Identification of Nonlinear Systems via Reservoir Liftings" is out in IEEE Control Systems Letters and accepted for presentation at IEEE CDC 2026. Congrats to all coauthors!
01
Who We Are
NEAR is a robotics lab working at the intersection of bio-inspired robotics, dynamics, control, and machine learning. We ponder:
“How can a robot's body become part of its intelligence?”
02
What We Do
We study embodied intelligence across robot morphology, sensing, nonlinear dynamics, system identification, control, neuromorphic computing, online adaptation, and lifelong learning.
Embodied & Bio-Inspired Robots
Design dynamic robotic bodies whose morphology, mechanics, and sensing contribute directly to behavior.
Dynamics & System Identification
Understand nonlinear, coupled, time-varying, and underactuated physical systems through modeling and data.
Adaptive & Learning-Based Control
Develop controllers that combine physical models with learning and adaptation.
Neuromorphic & Lifelong Intelligence
Explore efficient neural computation and continual adaptation for robots operating under real-world constraints.
Flapping-wing robots are one of our flagship testbeds. Their aerodynamics, compliant structures, sensing, actuation, and control are inseparably coupled, making them ideal systems for studying embodied intelligence under real physical constraints.
03
How We Get There
We do not treat intelligence as software added after the robot is built. We co-design the body and the brain.
Create robotic bodies whose morphology, mechanics, actuation, and sensing are designed together.
Model and identify the nonlinear physical dynamics governing their behavior.
Use data-driven, neural, and neuromorphic methods to extract useful structure from interaction.
Close the loop on physical robots so behavior can change with uncertainty, new environments, and accumulated experience.
From mechanics to mathematics, and from neural computation to real-world experiments, NEAR works across the complete loop of embodied intelligence.
04
What We Value
Intelligence is embodied.
The body is not merely a plant to be controlled. Morphology and dynamics can shape sensing, computation, and behavior.
Learning meets physics.
We combine data-driven methods with physical models and domain knowledge rather than treating robots as black boxes.
Robots should keep adapting.
Real environments change. Robots should improve and adapt beyond their original training conditions.
Ideas must survive reality.
Simulation is valuable, but physical experiments are the ultimate test.
Breakthroughs cross boundaries.
Challenging robotics problems live between mechanics, control, machine learning, neuroscience, materials, and biology.
05
Toward machines that adapt as naturally as living systems.
Our long-term goal is to create robots that can sense through their bodies, exploit their own dynamics, learn from experience, and continually adapt to the world around them while remaining efficient, interpretable, and trustworthy.
Intelligence should not be confined to an algorithm running inside a machine. It should emerge from the machine as a whole.
NEAR the future of embodied intelligence.