Langjing Xu

Graduate Research Project – Indoor Localization using Deep Learning

• Developed a Wi-Fi fingerprinting-based indoor localization system using multi-task CNN to predict building, floor, and (X, Y) coordinates.
• Designed architecture combining classification and regression tasks with weighted loss functions for robust performance.
• Compared Baseline CNN, Deep CNN, and ResNet CNN, achieving stable results on real RSSI datasets in complex indoor environments.
• Focused on model design, training stability, and performance analysis under varying signal conditions.

Course Project – Multi-Agent Decision-Making in Autonomous Driving

• Designed a reinforcement learning-based model simulating cooperative and competitive behavior among autonomous agents.
• Integrated deep learning behavior prediction into a dynamic game-theoretic framework.
• Applied to simulated highway merging scenarios with real-time strategy updates.

Course Project – Adaptive Buck-Boost Converter with Feedback Control

• Modeled a digitally controlled Buck-Boost converter with MATLAB/Simulink.
• Implemented closed-loop feedback control for voltage stabilization.
• Verified stability and control performance through hardware simulations.