Xindong Lin

Hi, I'm in the final year of my MSc in Computer Science at UBC, where I work with Professor Michiel van de Panne on physics-based character animation and robotics. My research focus is on planning and control for the skilled movement of physics-based humanoid characters, building on deep reinforcement learning methods. I'm also broadly interested in embodied AI and computer graphics. My ultimate goal is to create universal intelligent agents that can replace humans for dangerous tasks (or agents which surpass human capabilities😈). Previously during my undergrad, I worked with Professor Dinesh K. Pai on motion capture topics.

Prior to the above, I worked in the industry as a software engineer, where I developed and maintained Web Apps with JavaScript and Python. During my leisure time, I'm active in a variety of sports and play competitive FPS games such as Valorant, CS:GO and Apex Legends. Fun fact: I'm an Immortal (NA top 0.5%) player in Valorant.

Update: I'll be graduating in May 2026 and am actively looking for full-time opportunities related to robotics, character animation, embodied AI, or software engineering. If you are interested in my profile, please feel free to contact me via email.

More about me down below! 😄

News

Nov 2025

Mini-Pi-Plus with MimicKit

I created a demo of Mini-Pi-Plus using Xue Bin's new codebase MimicKit. Check out the link for more details! Mini-Pi-Plus on the right.

Education

Publications

SIGGRAPH Asia 2024

Measuring Human Motion Under Clothing

Luis Bolanos, Pearson Wyder-Hodge, Xindong Lin, Dinesh K Pai

Paper

This paper proposes and evaluates a system, EMob, for estimating body motion under clothing using a combination of very small electromagnetic sensors attached directly to the skin and traditional optical sensors on the clothing.

Projects

Learning to plan in bouldering (Ongoing)

RLIsaacLabPlanning

This project trys to learn a high-level planner that accounts for both the capability of the low-level controller and the hand hold pattern to solve hard bouldering problems. The picture is a toy example on a donut wall. More updates incoming!

GPU-accelerated Humanoid Stepping

IsaacLabRL

Code

This project reimplements the ALLSTEPS paper in IsaacLab. It analyzes the performance of ALLSTEPS on different simulations and optimizes the reward function to improve the performance of the humanoid robot in simulation. More updates are coming soon.

Robot2robot: Enhance imitation trajectory transfer with reinforcement learning

PythonC++RaiSimRL

CodeReport

This project uses RL to adapt open loop imitation trajectories from a source robot to a target robot with different degrees of freedom (DOF).

Mesh-based geometry processing algorithms implementation

C++OpenGL

CodeReport

Basic mesh-based geometry processing algorithms implemented in C++ and OpenGL.

Local smoothness layer for implicit coordinate learning

Report

This project aims to reconstruct 1D/2D/3D signals using neural networks, with an adaptive focus (LS-Layer) on different levels of detail in different regions.

Experience