Graduate Projects in UT Austin
Ongoing... (wireless tracking and imaging)
Undergraduate Projects in SJTU
Research on Indoor Localization
Temporal Correlation of RSS Improves Accuracy of Fingerprinting Localization
This work mainly focus on studying the temporal correlation effects on RSS fingerprinting based indoor localization method. From theoretical modeling and formula derivation, we found that temporal correlation of signals can definitely correct the localization criteria for MLE method, so that it can significantly improve the fundamental limits like accuracy and reliability. With expected experiment results, the paper was accepted by INFOCOM 2016.
Foxconn Cooperation Project
Posiiot Location Based Service System in iOS Development Part
We build a indoor positioning system named as "Posiiot". The whole system includes iOS/Android smartphone clients, website and server. I lead my team members to develop an iOS localization application for Foxconn. The iOS development consists of the BLE scanning, Map displaying, Pedometer, Info and Communication components. We also designed and implemented the location determination algorithms, both online k Nearest Neighborhood for Wi-Fi RSS fingerprints and online gradient descent method for Bluetooth RSS. The accuracy of our system can be within 2 meters.
Ericsson Cooperation Project
Dallas Load Testing Tool to Simulate Real User and Network Behaviors
During the second half of last year, I assisted in a cooperation project between IWCT (Institute of Wireless Communication and Technology) andEricsson. I renovated the traffic model and wrote a simulation software by C++ to model the user behavior propagation in 3GPP communication network. In particular, I read tens of thousands of Erlang codes to build the traffic model as a state machine for user activities in WCDMA network, as well as wrote a simulation software to model the user behaviors of the cellular network with error within 0.1%. In addition, I also simulated the traffic packages and user activity transitions in MATLAB to prove its stability.
Intelligent Transportation System
Lane-level Localization System Leveraging Sensors in Smartphone
We built a crowdsourcing based lane localization system to identify the number of lanes in a road and locate which lane the vehicles are in real time, so that it can facilitate the travel programming of the pilotless automobile and high precision vehicle navigation. The key idea of the system is leveraging the drivers’ smartphones to collect information from embedded GPS module, accelerator and gyroscope sensors, which are then fused by Integrated Multiple Model (IMM), and finally, identify the lane-level locations with constraint k-means clustering algorithm in server side.