M3LSDVRP

Multiobjective Solver with ML Models for routing and packing.

This is the code for paper "Machine Learning Assisted Multiobjective Evolutionary Algorithm for Routing and Packing" online version out soon

Multi-objective split delivery vehicle routing problem with 3-dimensional loading constraints (M3LSDVRP)

Framework of multiobjective routing packing solver

1. Installation:

   • Download

   • Compile a jar file named [name.jar] with RPSolver.java as the entrance

2. Usage: Examples and guidelines on how to use the project.

   java -jar name.jar ./ 4 2 1 60 0.4 1

3. Parameters:

   arg 0 : path of inputs and outputs and ML models (default current path "./")

   arg 1: type of ML models (default 4)

   arg 2: type of packing heuristic (default 2)

   arg 3: number of repeat, invalid, maximum running time is set to be the stopping criteria (default 1)

   arg 4: maximum running time (default 60)

   arg 5: theta (default 0.4)

   arg 6: encoding strategy (default 1)

   	String input_directory = args[0]+"/data/inputs"; 
   	String output_directory = args[0]+"/data/outputs";	
   	String model_directory = args[0]+"/model/"; 
   	classifier_type = Integer.valueOf(args[1]).intValue(); 	
   	// type of classifier: 0=linear, 1=MNB, 2=logistic regression, 3=svm_rbf, 4=CART, 5=random forest 10,6=random forest 20,7=random forest 50
   	packing_type = Integer.valueOf(args[2]).intValue();
   	// type of packing: 0=simpleHeuristic, 1=simpleHeuristic+newHeuristic1, 2=simpleHeuristic+newHeuristic1+newHeuristic2	
   	n_repeat = Integer.valueOf(args[3]).intValue();
   	n_repeat = 10000; 
   	// number of repeats: 1	
   	TIME_LIMITS = Integer.valueOf(args[4]).intValue();
   	// time limit: 60
   	theta = Double.valueOf(args[5]);
   	// theta: 0.2,0.3,0.4,0.5,0.6,0.7
   	encoding_strategy = Integer.valueOf(args[6]).intValue();
   	// encoding strategies: 1: full, 2, 3, 4

4. Results:

5. Visualization:

   An online visualization platform is provided by HUAWEI in https://www.noahlab.com.hk/logistics-ranking/#/simulator

   The results can be loaded in the platform for visualization and analysis

6. License: MIT