diseño de un dss para resolver el problema de ruteo
DESCRIPTION
Resumen del proyecto de investigación expuesto en la ALIO-INFORMS Joint International Meeting 2010, en Buenos Aires. James Tomalá Robles, docente de la Universidad Tecnológica Equinoccial, extensión Salinas, expuso la contribución “Route Planning Software and Hybrid Genetic Algorithm”, en el grupo “Transportation and Logistics II”; cuya sesión incluyó también la participación del profesor Tsutomu Suzuki de la Universidad de Tsukuba- Japón y Rudinei Luiz Bogo de la UFPR, Curitiba-Brazil.TRANSCRIPT
Route Planning Software and Hybrid Genetic Algorithm Design of a DSS to solve the routing problem in a Courier Service
Ph. D. Walter Vaca Arellano (EPN)
Ing. James Tomalá Robles
(UTE Universidad Tecnológica Equinoccial)
Ing. Johnny Pincay Villa (ESPOL)
EcuadorALIO-INFORMS meeting Buenos Aires
2010
Instituto de Ciencias Matemáticas-ESPOL
Motivation
In Ecuador, there are approximately 800 courier agencies, and all of them use empirical methods to
plan their routes
It directly affects the two objectives of integrated
logistics
Problem: the absence of a decision support system applying heuristic procedures for the transport operations planning of a courier company
Objectives1. Obtain a model
for courier delivery problem.
2. Design and develop a Metaheuristic based on genetic algorithm.
3. Propose a DSS design that uses the developed metaheuristic.
Mathematic formulationCapacitated Vehicle Routing Problem with smooth Time windows
Visitar todos los clientes una sola vez.
Todos los vehículos salen del depósito.
Mathematic formulation
Cada ruta es realizada por un solo vehículo.
La demanda no puede superar la capacidad del vehículo.
Se respeta la atención más temprana del cliente y se permite atraso.
Continuidad del tiempo y eliminación de subtours. Miller, Tucker y Zemlin [2].
Var. Binaria y positivos
GA for CVRPTW
Evolutionary Strategy
local search heuristics
Restrictions of time windows
GA for CVRPTW
Thangiagh [16], Bonrostro, Zhu[17], Homberger y Gehring [23]
On a review of GA publications, we may be concluded that GA requires modification of the classic genetic operators such as:
a) Redesign of the mutation and crossover operators.b) Inclusion of local search to improve solutions.c)And considerations of time constrains
Generation of population
Reproductive And
improving stage
BEGIN /* Hybrid Genetic Algorithm*/
Cargar_datos() //reads data from a filet←0
Po←generacion_poblacion_incial()
WHILE (t ≤ NUM_ITERACIONES) DO
/* Produce new generation*/
evaluacion_poblacion(Pt)
Pt<-mutacion_padres(Pt)
Pt← generar_hijos(Pt) // Produce new individuals with strategy (µ, λ )-EA t←t+1
END
END
Half - insertion heuristicHalf- Ramdomly
improving routesunifies routes
Replace if the child is better than one parent
The metaheuristic
chromosome representation
• chromosome:
1
0Dep
7
3
5
8
4
6
2
2
34
3
11
1
2
2
1
2
2 8 3 9 7 4 6 10 5 1
It adopts the permutation representation of integers, where the routes separator is a number greater than n, where n is the number of clients
evaluation , Fitness, Selection
• cost of the route:
• Penalty for delay :
• Cost of the solution (Fitness):
• Selection:
ordering the individuals in the population according to their fitness, that is, lowest to highest cost, and randomly selects among which are located below from 40th percentile.
Mutation
improve routes
unify routes
For i=1 ; i<= MOV_MUT
Select a local search operator{2-Opt *, relocation , exchange }Operator is applied
It tries to delete a route
Based on Homberger y Gehring [16].
combination operator
fatherBest routes
motherBest routes
Child
Inherit
The best
Based on uniform Crossover (UC) Áslaug Sóley Bjarnadóttir [11]
The combination strategy selects the best routes from the parents and insert them to the child provided it’s not conflict.
Proposal DSS
BD: Orders,demand
the TMS (Transportation managemen system)
WEB APPLICATION:DSS
planningmodule
APIGoogle Maps
component that calculates the distances between each pair of
customersInternet Hybrid GA
(Metaheurística )
Road and georeferential
DATA
Test results of solomon
optimum our solution
InstanceNumber of routes
Cost Number of routes
Cost
C101 10 827.30 10 828.94C102 10 827.30 10 828.94R104 10 982.010 10 1174.84R111 12 1048.70 11 1316.00RC103 11 1258.0 11 1424.34
The developed strategy to solve the problem requires less computational effort when data is grouped by area, it was found that in these cases, the number of iterations needed to reach a good solution is less than 40. On the other hand, we must increase the number of iterations when customers are completely randomly distributed in a geographical region.
Results – Study Case
RUT
A ORDEN DE VISITAS
CLIENTE
S
1
0-5-33-66-39-20-45-26-32-48-61-34-17-9-21-27-10-56-4-
12-29-0 20
2
0-7-38-37-23-57-24-35-52-25-54-65-51-60-43-62-36-63-
55-15-49-0 20
3 0-14-44-18-47-22-13-16-53-40-31-30-59-3-46-6-58-0 16
4 0-1-50-19-28-11-8-42-41-2-64-0 10RUTAS T ESPERA T ATRASO T SERVICIO T RUTA TOTAL
1 0:22:12 0:00:00 2:31:48 2:10:12 5:04:12
2 0:10:48 0:00:36 2:11:24 1:57:36 4:20:24
3 0:00:00 0:00:00 1:54:00 1:12:00 3:06:00
4 0:00:00 0:00:00 1:58:12 1:16:48 3:15:00
TOTAL 0:33:00 0:00:36 8:35:24 6:36:36 15:45:36
The developed Metaheuristic increases the level of service to 98.46%
According to company data, the average
service level of the first quarter of 2009, was 69.35%, ie, 20 clients were not treated on
time.
Example prototype using the Google Maps API
• available: www.ecualogistic.com/ruteo.php
The example prototype shows the solution of the case study.Each client has been located on the map according to their geographical location , additionally displays data such as order of visit, time of arrival and departure time. it was developed using javascript and dynamic language php.