Network Engineer.

Network

• Build a mesh network based on LoRa PHY with ESP32 board + LoRa SX1276 chip for low-cost smart agriculture.
• Work with MCCI LoRaWAN LMIC Library to realize the register level operating mode transition, interruption and data Tx Rx and CAD sequence of SX1276, and OS.Job level communication timing of ESP32.
• Design the LoRa protocol between gateway and multiple nodes. Propose strategies to ensure the auto-initialization and the long-term robustness of mesh network. Implement field test in Duke Smart Farm and simulation in NS-3.
• Investigate mesh for LoRa PHY network.

Cloud & ML

• Implement the data flow from sensors to microcontroller to cloud. Use Azure InfluxDB SaaS to do data fusion and ML.

Hardware

• Design and build the PCB for back-scatter tag. Choose components and design PCB layout.

Other

• Organize and standardize WiLOLoRa.lib, WiLOSensors.lib and WiLOPCB.lib to support the operating of LoRa mesh network, interfacing of multiple sensors, driving integrated circuit components and code reuse.
• Create KiCad schematic library and footprint library for PCB design.

School-Enterprise Joint Cultivation Program.

• Learned about server hardware structure and testing process.
• Built Log Server using Docker container technology to make client nodes send log information to a server node and render the log information divided by day, stored by client IP and accessed through the browser.
• Outcome: Project. [link]

Implement the query-based network monitoring application.

• Read TXT, PNG and WAV to bit-stream. Encode with (7,4) Hamming Code and (n,k) Systematic Cyclic Code.
• Pass Binary Symmetric Channel. Decode with Syndrome and Error Trapping Decoder. Reconstruct file then analyze.
• Outcome: Report. Location: LaTeX/report.pdf [link]

Implement the query-based network monitoring application.

• Create data generator with data generating thread and control receiving thread, stream processor with data receiving and processing thread and control receiving thread, database and website UI with function of controlling system and visualizing results.
• Design the socket communication between data generator and stream processor, HTTP communication between database and other parts, control the data generator and stream processor through website UI with websocket.
• Outcome: Report. Location: Project/Report/E6889 Stream Project Report.pdf [link]

Object tracking with Robust PCA initialization and sparse recovery.

• We designed a system (Clear Focus) which implements robust PCA to get the initial box automatically as the input for L1 tracker to remove the human interaction. To make the whole process faster while keep a high-resolution video, we also introduce super-resolution reconstruction of the video frame to the system. The Clear Focus system is tested on a sport video and daily scenario.
• Outcome: Report. Location: Project/Report/E6876_Report_cy2540_yl4189.pdf [link]

Team Leader.

• Program ESP8266 with MicroPython and C to deal with analog data, SPI/I2C bus communication and interrupts, to work with Google Geolocation, OpenWeatherMap and Twitter APIs, to run as an embedded server and receive Android App message, to communicate with MongoDB and do gesture recognition.
• Design a long-range, low-power, low-cost soil condition sensing system to simultaneously measure soil moisture and temperature over long duration at multiple locations. It is a distributed, scalable sensing network: using one LoRa gateway to link multiple sensor units via LoRa wireless communication. Data is stored in MongoDB. A dashboard website is built for spatial and temporal data visualization and systems control.
• Outcome: Embedded System and Website. A ESP8266 smart watch, a soil condition sensing system and a website. [link]

• Reproduce the conclusions with uniform samples in "Bandlimited Field Reconstruction for Wireless Sensor Networks".
• Experiment on Gaussian distribution of sampling and discuss the difference. Point out the problem of insufficient evidence for a conclusion in the paper.
• Outcome: Report. Location: Project/Paper & Presentation/RMT_Report_ChenyeYang_cy2540.pdf [link]

• Reproduce the behavioral biometric methods results about identification and authentication in "Smartphone and Smartwatch-Based Biometrics Using Activities of Daily Living" with "WISDM Smartphone and Smartwatch Activity and Biometrics Dataset" using classification methods like Decision Tree, Random Forest and K-NN.
• Conduct another method, Support Vector Machine, on the same dataset and compare the results of different methods applied to this specific problem and dataset.
• Outcome: Report. Location: Project/Report/StatL_Report_cy2540_bw2632_zx2269.pdf [link]

Team Leader.

• Built peripheral circuit and car structure based on myRIO development board, and programmed on myRIO with LabVIEW software.
• Manually realized a remote motion control of vehicle to enable it to go straight, swerve, back up, etc.
• Identified vehicle’s location by the infrared sensor and realized automatic obstacles avoidance.
• Outcome: Project. A toy car which can be remotely controlled or run spontaneously to bypass obstacles.

• Implemented Fixed-increment Batch Perceptron algorithm, Fixed-increment Single-sample Perceptron algorithm and Widrow-Hoff algorithm in Matlab.
• Realized real time face detection captured by camera based on OpenCV and using the Python programming language.
• Built face recognition model based on CNN, training classifier and identified human face in real-time camera data.
• Outcome: Project. [link]