Pixel Learning with Deep-RL for Mixed Autonomy Traffic (Advisor: Prof. Alexandre Bayen)

• Deploy the self-driving algorithm of FLOW, an open-source architecture integrating deep-RL with microsimulation tools, on both robotic vehicles and real vehicles.

• Utilize depth camera and web camera to realize 3D reconstruction of the 2D map.

• Use thresholding and K-means clustering to perform robotic vehicle detection in the 2D perspective.

• Utilize OpenCV to do green ball detection and tracking to realize the perception and localization of robotic vehicles in the 3D perspective.

The Mathematics of Poker-like Games (Advisor: Prof. A.J. Hildebrand)

• Implemented a 2v2 poker game prototype to demonstrate the McAdams Poker Model using Python.

• Developed the generalized McAdams Model, determined the best response strategy and the Nash Equilibrium Strategies, and investigated the best response iteration and its cycle length in this model.

• Presented the new model and results at UIC Undergraduate Mathematics Symposium.

• Led discussion sections with 30 students and taught Java and concepts of object-oriented programming.

• A total of more than 60 hours to help students during office hours, review labs, and answer questions.