Yunsheng Tian

Yunsheng Tian

PhD Candidate @ MIT CSAIL

yunsheng@csail.mit.edu

I am a final-year PhD candidate at MIT CSAIL advised by Prof. Wojciech Matusik, pushing the boundaries of robotics and scientific discovery. I earned my bachelor’s degree from Nankai University, and have interned at MIT-IBM Watson AI Lab, Autodesk Research, HKU, and MSR Asia. My research focuses on:

  • 🛠 Scalable robotic assembly of general objects: Industrial assembly lines are rigid, task-specific, and require significant human effort. I develop adaptive robotic systems that enable autonomous, flexible, and scalable assembly of complex products. My work spans task and motion planning, policy learning, physics-based simulation, and real hardware implementation.

  • 🔬 Accelerating experimental science via efficient optimization: Scientific discovery is often constrained by costly and lengthy experiments. I develop foundational and classical optimizers that combine probabilistic modeling with principled exploration to discover better physical structures, materials, robot designs, controllers, and beyond, with minimal experiments.

I am on the job market for research scientist/engineer positions tackling ambitious challenges.

Interests

  • 🤖

    Robotics

    Planning, Manipulation, RL

  • 🧠

    Machine Learning

    BayesOpt, Foundation Models

  • 🎨

    Computer Graphics

    Physics Simulation, CAD

Education

  • PhD in Computer Science, 2019 - Present

    Massachusetts Institute of Technology

  • MS in Computer Science, 2019 - 2021

    Massachusetts Institute of Technology

  • BEng in Software Engineering, 2015 - 2019

    Nankai University

Publication

ASAP: Automated Sequence Planning for Complex Robotic Assembly with Physical Feasibility

ASAP: Automated Sequence Planning for Complex Robotic Assembly with Physical Feasibility 🤖 🧠

International Conference on Robotics and Automation (ICRA) 2024
Robotic assembly plan generation for complex and general-shaped assemblies.
Boundary Exploration for Bayesian Optimization With Unknown Physical Constraints

Boundary Exploration for Bayesian Optimization With Unknown Physical Constraints 🧠

International Conference on Machine Learning (ICML) 2024
Efficient global optimizer for physical designs with unknown constraints.
Assemble Them All: Physics-Based Planning for Generalizable Assembly by Disassembly

Assemble Them All: Physics-Based Planning for Generalizable Assembly by Disassembly 🤖 🎨

ACM Transactions on Graphics (SIGGRAPH Asia) 2022
Physics makes motion planning for complex assemblies simple, fast, and generalizable.
JoinABLe: Learning Bottom-up Assembly of Parametric CAD Joints

JoinABLe: Learning Bottom-up Assembly of Parametric CAD Joints 🧠 🎨

Computer Vision and Pattern Recognition (CVPR) 2022
Learning to assemble on a big CAD assembly dataset approaching human-level accuracy.
Evolution Gym: A Large-Scale Benchmark for Evolving Soft Robots

Evolution Gym: A Large-Scale Benchmark for Evolving Soft Robots 🤖

Neural Information Processing Systems (NeurIPS) 2021
The first and largest benchmark for evolving soft robot morphology.
Diversity-Guided Multi-Objective Bayesian Optimization With Batch Evaluations

Diversity-Guided Multi-Objective Bayesian Optimization With Batch Evaluations 🧠

Neural Information Processing Systems (NeurIPS) 2020
Efficient multi-objective optimizer for expensive black-box functions.
Prediction-Guided Multi-Objective Reinforcement Learning for Continuous Robot Control

Prediction-Guided Multi-Objective Reinforcement Learning for Continuous Robot Control 🧠 🤖

International Conference on Machine Learning (ICML) 2020
Efficient multi-objective RL for discovering all Pareto-optimal policies.
Fluid Directed Rigid Body Control using Deep Reinforcement Learning

Fluid Directed Rigid Body Control using Deep Reinforcement Learning 🤖 🎨

ACM Transactions on Graphics (SIGGRAPH) 2018
Deep RL can control high-dimensional fluid-rigid coupling dynamics.

Preprint

AutODEx: Automated Optimal Design of Experiments Platform with Data- and Time-Efficient Multi-Objective Optimization

AutODEx: Automated Optimal Design of Experiments Platform with Data- and Time-Efficient Multi-Objective Optimization 🧠

Neural Information Processing Systems (NeurIPS) 2023 RealML Workshop
Easy-to-use experimental design platform based on asynchronous multi-objective BO.
Accelerated High-Entropy Alloys Discovery for Electrocatalysis via Robotic-Aided Active Learning

Accelerated High-Entropy Alloys Discovery for Electrocatalysis via Robotic-Aided Active Learning 🧠 🤖

Neural Information Processing Systems (NeurIPS) 2023 RealML Workshop
(Best Student Paper)
Accelerated scientific discovery through active learning and autonomous robots.

Project

Autonomous Robotic Assembly of a Skateboard Truck

Autonomous Robotic Assembly of a Skateboard Truck 🤖

Autodesk Research Project 2023
Autonomous assembly of a mechanical skateboard truck with a flexible robotic system.

Service

Conference Reviewer: SIGGRAPH/SIGGRAPH Asia, CoRL, ICRA, IROS, ICLR, ICML, NeurIPS, AISTATS

Journal Reviewer: RA-L, TVCG, TEVC, Soft Computing, TMLR

Workshop Organizer: CoRL 2024 Robotic Assembly Workshop

Invited Talks:

  • [2024/07] DigitalFUTURES Workshop: “Towards Generalizable and Adaptive Robotic Assembly of Industrial and Everyday Objects”
  • [2023/11] Tsinghua University IIIS Seminar: “Towards Automated Planning and Execution for Generalizable and Complex Robotic Assembly”
  • [2023/08] Autodesk Research Connections: “Physics-Based Planning for Automated Robotic Assembly”
  • [2023/03] MIT Graphics Seminar: “Physics-Based Planning for Automating Contact-Rich Assembly”
  • [2021/04] MIT CSAIL Alliances Tech Workshop: “Automated Multi-Objective Optimal Experiment Design Platform”
  • [2020/11] Lam Research Seminar: “Multi-Objective Bayesian Optimization for Adaptive Experimental Design”

Teaching and Mentoring:

  • [Fall 2023] Teaching Assistant for MIT EECS Advanced Computer Graphics/Computational Design and Fabrication
  • [Fall 2020 - Present] Research Mentor of 8 students participating in MIT UROP (Undergraduate Research Opportunities Program)