Project

Optimal Control

Optimal Control Solver - New Miser(Supervisor:Changjun Yu in Shanghai University)

2020.9-2023.7
We want to rebulid a classical optimal control solver- Miser3.2(2003) via Julia, which can make it new and fast. What's more, we are expected to add some new technique such as time scaling and auto-gradient.

PDE Optimal Control(Supervisor:Changjun Yu in Shanghai University)

Recently we have finished a paper about using exact uzawa method to solve elliptic optimal control problems with its dual form. It will be submitted soon.

Integer Programming

ML4CO
Machine Learning for Combination Optimization(Supervisor:Yuhong Dai in Chinese Academy of Sciences)

2021.6-2022.4
We want to make connections between Integer Programming Solver and Reinforcement Learning/meta learning.
For some types of LP problems, we give some new pivot choosing rule to Simplex method via meta-learning or other reinforcement learning methods.
Learn some basic knowledge about integer programming and inplement the Bland's rule and the steepest edge rule on the VRP problem.

Optimize Beijing's Public Transportation System Network(Supervisor: Yuan Sun in The University of Melbourne)

2022.4-2022.10
Based on the existing bus and subway lines in Beijing, we first detect the abnormal situation in the transportation network, and then extract the problem station after dividing the abnormal area. and combine the Orienteering Problem(OP) and the origin-destination(OD) matrix to optimize the existing bus lines in Beijing. Then we have real experiments in Tongzhou District, Beijing.

Online optimization of a dial-and-ride problem with the integral primal simplex(Supervisor:Antoine Legrain in Polytechnique Montreal)

2022.9-2022.12
We will develop a simulation tool in python to test the code. We will not only code dedicated algorithms to compute compatibility matrices. but also develop basic algorithms (as a greedy) for the DARP in NYC. If time permitted, we will develop visualization tools.