ShanghaiTech University / School of Information Science and Technology / STAR Center / Model-free Autonomous Agent and Intelligent Control Lab

About Us

Welcome to the Model-free Autonomous Agent and Intelligent Control (MAgIC) Lab at ShanghaiTech University. Our research spans the development of bio-inspired underwater and aerial robots—such as robotic fish and flapping-wing vehicles—with a focus on perception, planning, control, and multi-agent coordination. Currently, the core of our work lies at advanced motion control for robotic systems operating in dynamic fluid environments.

Underlying these applications is our fundamental research into the modeling and control of complex, nonlinear, time-varying systems across multiple scales. We develop adaptive, data-driven, and learning-based control theories with provable guarantees, tackling challenges such as uncertainty, disturbance, and system constraints.

Led by Prof. Yang Wang, our team consists of passionate PhD and master students dedicated to pushing the boundaries of fluidic-interactive robotic systems and intelligent control theory. We welcome collaborations and discussions with researchers who share our interests.

News

phd-graduation

14 May 2026

A Historic Milestone for MAgIC Lab: Congratulations to Heng Zhang, Our First Ph.D. Graduate!

MAgIC Lab warmly congratulates Ph.D. student Heng Zhang on successfully completing the doctoral thesis defense and becoming the first Ph.D. graduate in the history of the lab. Heng Zhang conducted doctoral research under the supervision of Prof. Yang Wang and made significant contributions to the lab's research activities during the Ph.D. program.

As the first doctoral graduate of MAgIC Lab, Heng Zhang's achievement marks an important milestone in the development of the lab. We sincerely wish Heng Zhang continued success in future academic and professional endeavors.

IFAC

14 Apr 2026

Congratulations to MAgIC Lab for two papers accepted by IFAC2026.

The papers entitled 'Switching-based Adaptive Feedforward Control for Uncertain Linear Multivariable Systems: Periodic Disturbance Cancellation' and 'Model Reference Adaptive Control without High-Frequency Gain Knowledge via Derivative Injection and Global HOSM Differentiators' have been accepted for presentation at the 23rd IFAC World Congress. The former proposes a switching-based adaptive feedforward approach to reject multi-sinusoidal disturbances in uncertain MIMO systems, while the latter presents a derivative-injection and HOSM differentiator enhanced MRAC scheme to achieve stable tracking for arbitrary relative-degree plants with unknown high-frequency gain.

ICRA

31 Jan 2026

Congratulations to MAgIC Lab for the paper accepted by ICRA2026.

The paper entitled 'Agile and Controllable Omnidirectional Fast-start Maneuvers of Robotic Fish via Bio-inspired Reinforcement Learning' has been accepted for publication in IEEE ICRA. This paper presents a deep reinforcement learning framework for multi-joint robotic fish to reproduce biologically inspired C-start fast-start maneuvers under highly unsteady fluid dynamics. By embedding key biological features into the learning design and training in a high-fidelity CFD environment, the method enables robotic fish to autonomously discover effective launch strategies and achieve controllable high-acceleration motions with improved speed and maneuverability.

ICRA

31 Jan 2026

Congratulations to MAgIC Lab for the paper accepted by ICRA2026.

The paper entitled 'Deep Photonic Reservoir Computing for Ultra-Fast Feedforward Dynamic Compensation of UAVs in Confined Environments' has been accepted for publication in IEEE ICRA. This paper presents an ultra-fast deep photonic reservoir computer (PRC)-based feedforward compensation framework to mitigate nonlinear, time-varying proximity-induced residual forces (e.g., ground/ceiling effects and wake recirculation) that degrade UAV tracking in narrow, cluttered spaces, achieving millisecond-level training via linear ridge regression and ultra-low-latency inference while improving closed-loop stability under high-fidelity CFD evaluations.

View All Events →