Shiyu Wu

A Learner. A System Builder.

Personal Information


  • Shiyu Wu / 吴矢瑀
  • Allen Wu
  • Shanghai, China
  • Austin, TX

Contact Information


Skills


About Me


  • Hi All! I am Shiyu Wu. I also go by my English name, Allen Wu.

  • I was born in Shanghai, and spend most of my lifetime in this amazing city!

  • After spending two wonderful years at Ann Arbor, I've moved to Austin!

  • I got my dual bachelor degrees from both University of Michigan, and Shanghai Jiao Tong University.

  • I have broad interests in EECS field, particularly Computer System, and Robotics.

Education Background


  • University of Texas at Austin, College of Natural Sciences

    M.S. in Computer Science, GPA: N/A.

  • University of Michigan - Ann Arbor, College of Engineering

    B.S.E. in Computer Science, GPA: 3.97/4.0, Summa Cum Laude.

  • Shanghai Jiao Tong University, UM-SJTU Joint Institute

    B.S.E. in Electrical and Computer Engineering, GPA: 3.75/4.0, Outstanding Graduate.

Notable Projects


  • [Project1] Learning Behavior Trees from Demonstration Demo

    • Our Paper: Kevin French, Shiyu Wu, Tianyang Pan, Zheming Zhou, Odest Chadwicke Jenkins, "Learning Behavior Trees from Demonstration" Accepted to IEEE International Conference on Robotics and Automation (ICRA) 2019
    • [Goal] In order to allow the end user to teach new tasks to robots without expert knowledge, we propose a new learning from demonstration pipeline that incorporates Behavior Trees, a control architecture, as a new form of policy produced by learning from demonstration.
      [Advisor] Prof.Chad Jenkins
      [Contribution] I designed and implemented BT-Espresso algorithm, which is an algorithm converting the learned decision tree from demonstration into a Behavior Tree, and other related algorithms. More specifically, I incorporated several boolean simplification algorithms into the system, including Espresso heuristic logic minimizer, tabular method, which could efficiently prune the behavior tree without performance loss. Further improved the system using multi-threading and multi-process methods. Developed project GUI based on PyQt module.
      [Demo]

  • [Project 2] GlassLoc: Plenoptic Grasp Pose Detection in Transparent Clutter Demo

    • Our paper: Zheming Zhou, Tianyang Pan, Shiyu Wu, Haonan Chang, Odest Chadwicke Jenkins, "GlassLoc: Plenoptic Grasp Pose Detection in Transparent Clutter" Submitted to IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2019)
    • [Goal] Detect grasp poses of translucent objects or objects under transluceny clutter using plenoptic camera and learning methods.
      [Advisor] Prof.Chad Jenkins
      [Contribution] I Built the pipeline of incorporating Grasp Pose Generator (GPG) to extract grasp poses’ features of translucent objects. Implemented data augmentation methods to generate training data. Collected light field images and point cloud training data. Migrated PointCNN for feature learning on depth likelihood volumes.
      [Demo]

  • [Project 3] Chez Homie - Mobile Manipulation Project Demo

    • On going
    • [Goal] Chez Homie is a mobile manipulation project targetting on enabling robot to perform automatic food delivery in Bob and Bettey Beyster Building, developed by Progress Lab at the University of Michigan.
      [Advisor] Prof.Chad Jenkins
      [Contribution] I designed the state manager of the robot by incorporating Behavior Trees and its common interfaces between Robot Operating System (ROS), based on open-source package. Developed human-robot interactions features, including text speech conversion system, head actuation, robot contorl user interfaces, and etc.
      [Demo]

  • [Project 4] Asciiii - Ascii Style Converter Demo

    • [Goal] Asciiii is an ASCII style converter made during MHacks11. It supports image inputs and ouputs the ASCII-style text strings depicted by the edge information of the input.
      [Online Website] Asciiii
      [Contribution] I designed and implemented the image processing algorithms based on OpenCV library. Incorporated Canny edge detection, hamming distance-based grid matching, and other algorithms. Improved the system using heuristic algorithms and multi-threading to support video streaming application.
      [Demo]

  • [Project 5] Gun Violence Study Demo

    • [Goal] This project focuses on predicting future gun violence cases by using machine learning methods.
      [Advisor] Prof.Barzan Mozafari
      [Contribution] I collected and formulated shooting incident, gun ownership, and other gun related data. Used Hidden Markov Model and ARIMA time-series model to predict the potential location and profile of the shooter of next shooting incident. Performed statistical anaylsis, including ANOVA, false discovery rate analysis.
  • [Project 6] Music Genre Classification Demo

    • [Goal] This project focuses on classifying music genres by applying digital signal process and machine learning techniques.
      [Advisor] Prof.Mert Pilanci
      [Contribution] I applied FFT and mel-frequency cepstrum to audio signals to extract features of different music genres. Apply machine leaning algorithm, including k-Nearest Neighbors, k-Means, Multi-Class Support Vector Machine, and Random Forest to perform genre classification. Achieve approximate 85% correctness on testing dataset. Designed and deployed the project website. This project got the highest score among all course projects.
      [Demo] More information could be found on our project website! Click here.
  • [Project 7] Paperman-Automatic Paper Sorting Machine Demo

    • [Goal] Paperman is designed to automatically sort papers based on its usage over two sides.
      [Advisor] Prof.Mian Li
      [Contribution] I implemented the algorithm on Arduino to control sensors and step motor for paper transmission and classification. Designed and constructed mechanical structures of the machine to transmit paper. Used CAD to draw the 2D blueprints, 3D images, and concept diagrams. The machine had averaging 90 percent rate of correct sorting. The project was honoured with Best Technical Achievement Award in the design expo.
      [Concept Diagram]

Other Experience


  • Teaching Assistant: MKT382 Data Analysis & Visualization

    • Instruct students with basic knowledge of R, SQL, and Python.
  • Teaching Assistant: EECS484 Database Management System

    • Worked with Prof. H.V. Jagadish, Prof. Atul Prakash
    • Design and implement one subpart, Grace Hash Join, of Project 4. Regular duty includes holding weekly office hour, discussion and review session, homework and exam making up and grading.
    • Student Evaluation: 4.2/5
  • Modi Education - High School Tutor

    • Part time job as an tutor for high school's maths and physics courses.
    • Duty: 8-hours' weekly lectures. Design specific homework.

Honors / Awards


  • National Scholarship

  • King, Roger Scholarship

  • EECS Scholar

  • Jun Yuan Scholarship (Two Times)

  • UM-SJTU JI Distinguished Academic Achievement Award

Personal Interests


  • I am a big sports fan. I enjoy basketball, billiards, tennis, Formula one. Golden State Warriors point guard Klay Thompson is my favourite basketball star!

  • I love reading a lot, especially philosophy-related and detective novels. My favorite writers include Agatha Christie, Albert Camus, and Hermann Hesse.

  • Contemporary art always fascinates me!