I am a roboticist specializing in motion planning and control for legged locomotion. I completed my Ph.D. at UCLA, where I worked with Prof. Dennis Hong. I am currently completing a postdoctoral position at the Georgia Institute of Technology, where I've had the opportunity to collaborate with Prof. Ye Zhao.
My first authored papers received the Best Paper Award on Safety, Security, and Rescue Robotics at IROS 2019 and was a finalist for the Best Paper Award at the UR 2024. I am a member of the world championship team for the RoboCup 2024 competition.
During my M.S. studies, my research focused on the hardware aspects of robotics, including the development of wall-climbing robots, grippers, and humanoids; notable works include SiLVIA and SCALER. During my Ph.D., I have shifted toward creating general robotics frameworks for motion planning and control. My research focuses on developing scalable mixed-integer convex programming algorithms for robotics. Building on my contributions, including logic network flow and Benders decomposition, I aim to scale mixed-integer convex programmings for robot motion planning, with applications in search and rescue, urban logistics, factory automation, and beyond.
We propose a hybrid motion planning and control framework based on Generalized Benders Decomposition that controls a cart-pole system with randomly moving soft-contact walls reaching speeds 2-3 times faster than Gurobi, oftentimes exceeding 1000Hz.
Xuan Lin, Jiming Ren, Samuel Coogan, Ye Zhao ICRA, 2025
|
paper /
code
We propose Logic Network Flow, an innovative optimization formulation for motion planning under temporal logic constraints. Synthesized with Dynamic Network Flow, our framework accelerates the computation by tightening the convex relaxations.
Xuan Lin, Gabriel Fernandez, Dennis Hong UR, 2024, finalist, Best Paper Award
|
paper /
code
We compare the data-driven performance of two MIBLP reformulations: mixed-integer programming (MIP) and mathematical programming with complementary constraints (MPCC). This evaluation is conducted on a book placement problem featuring discrete configuration switches and bilinear constraints.
Xuan Lin, Jiming Ren, Samuel Coogan, Ye Zhao Ongoing work, 2024
We demonstrate task and motion planning for time-critical search and rescue tasks using humanoid robot teams inside a realistic battlefield simulation environment using MuJoCo.
Yusuke Tanaka, Xuan Lin*, Yuki Shirai*, Alexander Schperberg, Hayato Kato, Alexander Swerdlow, Naoya Kumagai, Dennis Hong (*equal contribution) IROS, 2022
|
paper
We introduce SCALER, a quadrupedal robot capable of climbing bouldering walls, overhangs, and ceilings, as well as trotting on the ground while carrying payloads up to 233% of its weight on flat surfaces and 35% on vertical walls. The first author received the IROS 2022 SICE International Young Authors award. Congratulations!
Xuan Lin, Jingwen Zhang, Junjie Shen, Gabriel Fernandez, Dennis Hong IROS, 2019, Best Paper Award on Rescue Robotics
|
paper
We introduce SiLVIA, a hexapod robot that demonstrates climbing between two walls with bare foot and planned motion, the first robot to demonstrate such capability.
Accelerating Mixed-integer Convex Programs for Motion Planning and Control
One of my primary research areas is accelerating Mixed-Integer Convex Programs (MICPs) for robot motion planning and control. Historically, MICPs have been considered too slow for real-time robotic applications. However, exciting recent advancements are breaking this barrier. A notable example is the Graph-of-Convex-Sets, which designs trajectories around obstacles with higher quality compared to traditional sampling-based methods, demonstrating success in industrial robots.
My approach to accelerating MICPs for robot motion planning and control focuses on three general aspects: (1) designing MICP formulations with tight convex relaxations, such as logic network flow; (2) designing MICP model decompositions, using techniques like Benders decomposition and Branch-and-Benders Cut; and (3) utilizing and customizing commercial solvers, whose efficiency has significantly improved over the years with advances in computation power. The connections between these components are illustrated in the figure below.
