GE-DDRL-ROUTER
The source code for "GE-DDRL: Graph Embedding and DeepDistributional Reinforcement Learning for ReliableShortest Path: A Universal and Scale Free Solution".
Table of Contents
1.Requirements
numpy, cvxopt, scipy, networkx, gensim, pytorch and other common packages.
2.Framework of GE-DDRL-Router
2. How to use
step 1: Create a simulative environment(ENV), you must choose one transaction data or make a data by yourself, and specify one RSP objective. For some transaction data without the sigma of link travel time, you can generate them through the funtions in func_new.py, you can find the method you want in func_new.py or you can make your own method to generate sigma of link travel time;
step 2: Create a graph embedding module(GE) base on the environment that is created on step 1, some parameters are required to set;
step 3: Create a Agent based on the ENV and GE that are created on step 1 and step2;
step 4: Configure the parameter of the Agent that is created on step 3;
(Optional) step 5: Using dijkstra to pretrain the Agent(Imitation Learning), which is a warm start for DRL-Router.
step 6: Start the training of the Agent, we need to set the training parameters num_iterations, When the training is finished, we got a Policy. The more training times, the better performance of results will be.
3. Running
You can start your tranning in the main.py.