Project

Deep Q-Networks Reinforcement Learning on Super Mario Bros.™

Project, Advanced Topics in Machine Learning, Caltech

We present a deep learning model to successfully learn control policies from high-dimensional input data using reinforcement learning. The model is based on the idea of Deep Q-Network (DQN), with convolutional neural network trained by Q-learning algorithm, whose input is tile representation of the screen and output is a value estimation function. Also, replay buffer, target network and double Q-learning are applied to lower data dependency and approximate real gradiant descent. We applied our model to Super Mario Bros.™, and get some good preliminary results.

For our poster, please see here.

D-star Lite Motion Planning Algorithm Implementation on Robot Operating System (ROS)

Project, Autonomy, Caltech

In this project, we have successfully achieved dynamic path-planning using D*lite based on frontier explorer class and a modified global-planner plugin. We have shown our system is working both in simulation and on real Turtlebot, and it enables path replanning even in narrow space with complicated obstacles which cannot be seen at the beginning.

Full-Stack Web Application Design (Limited Information)

MathWorks® Intern Project

As a summer intern in MathWorks Inc, I developed and improved a full-stack web application project on MathWorks^® online services, with JavaScript (HTML and CSS) in front-end coding and Dojo/jQuery as framework and Java/MATLAB for back-end integration. My work involved rich experience in a full software development cycle, including Requirement Analysis, Functional Specification, UX & Structure Design, Unit Test, System Maintenance, Feedback Connection, and Project Closure.

PageRank Algorithm Implementation via MapReduce

Project, Networks: Structure Economics, Caltech

In this work, we implemented PageRank Algorithm in matrix form (with Python), and adapted to MapReduce programming model; applied optimization strategy such as top-k convergence, heapsort and extrapolation to accelerate computing; andr anked 2nd among 20 teams in Caltech PageRank competition “Rankmaniac” and beat the last year champion.

Computing & Operating System Projects

Project, Operating System, Caltech

In this set of projects, we dealt with OS problems in modern computing systems thru programming with low level C and x86-64 Assembly. Projects include: Run Length Coder, C OOP, error handling and garbage collection, Explicit Heap Allocator, Thread Scheduling Library, User-space Virtual Memory System Design, and CPU Cache Optimization.

White LED Free-space Optical Communications (Li-Fi) System

Project, Applied Optoelectronics Lab Class, Instructor: Dr. Kaiwei Wang, Zhejiang University

In this work, we used white LED as light source, two programmable micro control unit (MCU) as transmitter and receiver ends, and built up a free-space communication link with zero bit error rate, 10^5 Hz rate, and over 1m distance. Applying noise processing algorithm and dynamic monitoring method, we successfully realized text, audio and video files communication between both ends.

Polarized Phase Modulation System

Project, Precision Interferometry and Applications, Instructor: Dr. Dong Liu, Zhejiang University

Contrast adjustment of interference pattern

In this work, we set up a phase-polarized phase modulation system for sample surface detection. A automatic dynamic motor control was applied to the system to adjust the contrast of interference pattern. This system was able to test samples with low-reflectivity in an noisy environment.

Google Sites

Report abuse