Summer 2022 reasearch about ride sharing VRP
The two folders contain codes and experiment results of two directions of a ride-sharing problem.
The first direction is
Formulate VRP with time constraints as IP + Learning to cut, but VRP with time factors is a MIP problem which cannot be solved by Learning to cut. So, we only solve a matching problem with this RL method.
The second direction is
Construction RL, Pointer Network + Policy Gradient, with time reward function. We made experiments in a 4-request and a 10-request scale problems.