Anran Zhang

Hi, my name is Anran Zhang. I am currently studying Mechatronics and Robotics (M.Sc) at Technical University of Munich. I am enthastic about generalizable robot manipulation and energy-based control. Currently I am doing my master thesis at Smart Robotics Lab.

I grow with different robot systems and finished several research projects in KI. Fabrik under supervision of Prof. Sami Haddadin. Lately I finished a 6 month Internship within realtime system group Agile Robots AG , under supervision of Dr. Eugen Nuss.

I obtained my bachelor's degree in Jilin University in China, majoring in Automotive Engineering, where I had a great time in Formular Student Team.

Email  /  CV  /  Github

Research

I'm interested in generalizable robot manipulation, computer vision and control theory. Some papers are highlighted.

To enable robots to manipulate objects in a generalizable way, we aim to develop a framework that:
- recontruct scene from single-view image input at object-level.
- distill 2D features to 3D scene leveraging foundation models
- transfer manipulation policy across category with feature matching of the scene

We provided a robust framework for autonomous and adaptive robotic manipulation. The framework consists of tank energy estimation and unified force-impedance control for a plugin-and-play experience for safety guarantee to constant interactions with the environment.

Aiming at intuitive visuotactile exploration of unknown 3D curvatures, we propose Vision-Augmented Unified Force-Impedance Control (VA-UFIC).
- combining tactile and vision perception for a robust online contact monitoring system
- integrating contact alignment monitoring into VA-UFIC to adapt the robot's stiffness and regulate the force profile.

Other Projects

Implement close-loop dynamic simulation to accelerate the development of in-house control algorithms. Dynamics and kinematics of the robot is computed using Pinocchio.

Calibration according to Standard ISO 9283, gather data from laser tracker and robot, evaluate the repeatablity performance, and optimize the kinematic model of the robot.

Deployed control algorithms (PID and Model Predictive Control) on scaled F10th racecar in simulation and transformed onto race track. It was fun!