Professor · EECS · UC Irvine

Control, learning, and computation in networked systems.

Research on dynamical systems theory, machine learning, and computational neuroscience, from brain-inspired computing to multi-robot autonomy.

Department of Electrical Engineering & Computer Science
University of California, Irvine
Director, UCI Robotarium
01 / Research

The common thread: how networks compute, coordinate, and fail.

Learning & dynamical systems

Methods for modeling and controlling dynamical systems from data, using Koopman operators, diffusion-based planners, and learning-enabled controllers that come with stability guarantees.

Brain-inspired computing

Architectures for learning and computation built on dynamical systems theory, including oscillator networks, associative memory, and physical intelligence.

Robotics & autonomous systems

Multi-agent coordination, path planning, and learning-based control, tested on hardware in the UCI Robotarium.

Network neuroscience

Control theory, dynamical systems, and spectral methods applied to neural data: cortical oscillations, effective connectivity inferred from recordings, and how structure shapes brain dynamics.

Cyber-physical systems

Observability, resilience, and security of networked infrastructure such as power grids, transportation networks, industrial control, and autonomous vehicles.

New research direction

An attention mechanism with a unique attractor per query.

Attention realized as the relaxation of a coupled-oscillator network with pinned anchors. The architecture admits one attracting equilibrium per query, reached from all but a single starting point, so the answer does not depend on where the dynamics begin. That uniqueness is what makes oscillator attention trainable on a substrate that, by default, supports many equilibria.

Read the research page
UCI Robotarium

A facility for multi-robot experimentation, brain-inspired computing, and physical AI.

Open to researchers, students, and industry partners. A motion-tracked arena, 3D fabrication, and on-site compute, so an idea can go from theory to hardware in the same building.

Explore the facility
12
Vicon cameras
sub-mm tracking
60m²
Experimentation
arena
3+
3D printers
multi-material
GPU
On-site compute
cluster
03 / News

From the group

Nov 2026

New preprint on cooperative trajectory generation from single-agent demonstrations

Diana Laura Morales's work with Ruoyu Lin and Magnus Egerstedt shows how to generate cooperative trajectories for heterogeneous teams with diffusion models trained only on demonstrations of single agents. The preprint is on arXiv.

Oct 2026

Feature article on computing with oscillator networks submitted to IEEE Control Systems Magazine

With Ahmed Allibhoy, we submitted a tutorial on how synchronization in coupled oscillator networks can compute, covering combinatorial optimization, associative memory, and attention. The preprint is on arXiv.

Sep 2026

New preprint on multi-agent flow matching with guarantees

Ruoyu Lin's work with Magnus Egerstedt on DeGG-Flow, a framework that guides multi-agent generative models to satisfy hard coupled requirements, is now on arXiv.

Sep 2026

Attention by Synchronization published in TMLR

The oscillator-attention paper with Taosha Guo is out in Transactions on Machine Learning Research.

Aug 2026

Provably Safe Generative Sampling published in TMLR with J2C certification

Darshan Gadginmath and Ahmed Allibhoy's paper on constricting barrier functions is out in Transactions on Machine Learning Research, and received a J2C certification (top 10% of accepted papers).

Jul 2026

NSF award on control-theoretic foundations for trustworthy generative AI

NSF CMMI supports work on making generative AI safe, coordinated, and reliable in safety-critical engineering systems. Runs October 2026 to September 2029.