Research

Global planning algorithm development based on the dynamical programming and Dijkstra's algorithm
Local control algorithm development based on reinforcement learning and model predictive control

Global planner, color mapped by 3D cost-to-go

Local planner with the capability to avoid dynamical obstacles, color mapped by cost-to-go

Autonomous Parking with deep Koopman representation and sample-based model predictive control (MPC)

Graduate Researcher, supervised by Dr. Yiqiang Han

Clemson University, Clemson, SC, USA, Aug 2020 - Dec 2020

Implement the MPC controller with deep Koopman representation for the autonomous parking

Graduate Researcher, supervised by Dr. Yiqiang Han

Clemson University, Clemson, SC, USA, March 2020 - July 2020

Developed a novel computation-efficient path-planning algorithm for the quadrotor based on the A* and D* algorithms
Implemented the proposed method in both Gazebo and real-world environments

Graduate Researcher, supervised by Dr. Yiqiang Han and Dr. Umesh Vaidya

Clemson University, Clemson, SC, USA, Aug 2019 - July 2020

Developed a novel data-driven control framework to control the unknown dynamical systems, which is based on the Koopman operator theory, model predictive control (MPC), and deep neural network (DNN)

Graduate Researcher, supervised by Dr. Yiqiang Han

Clemson University, Clemson, SC, USA, May 2019 - Feb 2020

Deployed simultaneous localization and mapping (SLAM) technique for the RC car in both robot operation system (ROS) and real-world environments
Simulated the path planning and following algorithms in MATLAB
Verified the algorithm in Gazebo

Graduate Researcher, supervised by Dr. Yiqiang Han

Clemson University, Clemson, SC, USA, Oct 2018 - Jan 2019

Built a reinforcement learning model for low-level control in Gazebo using AWS online service
Built a convolutional neural network (CNN) model for objects and lane detection, which is trained on the Nvidia-dgx2 GPU cluster
Deployed the models on the RC car for real-world application