Viviane Gascon François Meunier UQTR INFORMS Healthcare 2011 1.
-
Upload
mervyn-horn -
Category
Documents
-
view
215 -
download
0
Transcript of Viviane Gascon François Meunier UQTR INFORMS Healthcare 2011 1.
Viviane GasconViviane GasconFrançois MeunierFrançois Meunier
UQTRUQTR
INFORMS Healthcare 2011INFORMS Healthcare 2011 1
A public medical clinic offers activities to elderly at a A public medical clinic offers activities to elderly at a daycare centerdaycare center
Users are transported to the daycare center by bus (2) Users are transported to the daycare center by bus (2) or by taxior by taxi
Costs of taxis have been increasing in the last few yearsCosts of taxis have been increasing in the last few years
QuestionsQuestions : : Is it possible to reduce the costs related to the use of Is it possible to reduce the costs related to the use of
taxis ?taxis ? What would be the impact on the buses routes ?What would be the impact on the buses routes ?
INFORMS Healthcare 2011INFORMS Healthcare 20112
Cost analysisCost analysis Determining new routesDetermining new routes Greedy algorithmGreedy algorithm Ant colony heuristicAnt colony heuristic InterfaceInterface
INFORMS Healthcare 2011INFORMS Healthcare 201133
Cost analysis: economy if reduction of total distanceCost analysis: economy if reduction of total distance◦ Bus 1Bus 1: $1.283 /km: $1.283 /km◦ Bus 2Bus 2: $1.68 /km: $1.68 /km
TaxisTaxisThe costs vary with the #km traveled and the #users in a taxi on The costs vary with the #km traveled and the #users in a taxi on
a given trip.a given trip.
Increase the #users traveled by bus if it costs less than Increase the #users traveled by bus if it costs less than by by taxistaxis Determine new routesDetermine new routes
INFORMS Healthcare 2011INFORMS Healthcare 20114
UsersUsers
type of mobility: on foot, wheelchair, walkertype of mobility: on foot, wheelchair, walker living in a nursing home or notliving in a nursing home or not
(they must come back to their nursing home for lunch)(they must come back to their nursing home for lunch) with or without cognitive problemswith or without cognitive problems
(users with cognitive problems must be in homogenous groups and stay (users with cognitive problems must be in homogenous groups and stay the whole day)the whole day)
visit frequency: once or twice a weekvisit frequency: once or twice a week length of visit: half day or whole daylength of visit: half day or whole day 48 users on average per day48 users on average per day
INFORMS Healthcare 2011INFORMS Healthcare 20115
Two busesTwo buses
each bus covers a sectoreach bus covers a sector bus capacity varies with users’ type of mobilitybus capacity varies with users’ type of mobility
(users with wheelchairs and walkers take more space reducing the bus (users with wheelchairs and walkers take more space reducing the bus capacity)capacity)
INFORMS Healthcare 2011INFORMS Healthcare 20116
Route constraintsRoute constraints
total duration must not exceed 60 minutestotal duration must not exceed 60 minutes(including traveling time and service time)(including traveling time and service time)
sector 1: three pick up and return routes per daysector 1: three pick up and return routes per day sector 2: two pick up and return routes from Monday to sector 2: two pick up and return routes from Monday to
Wednesday, one on Thursday and three on FridayWednesday, one on Thursday and three on Friday
INFORMS Healthcare 2011INFORMS Healthcare 20117
Position users on a map: two sectors are considered, Position users on a map: two sectors are considered, one for each busone for each bus
Define new clusters of users for each sector, for each Define new clusters of users for each sector, for each dayday
INFORMS Healthcare 2011INFORMS Healthcare 20118
Generate new routes Generate new routes ◦ keep same days of visit for users with cognitive problems keep same days of visit for users with cognitive problems
(Tuesday and Friday) and add some users actually traveled by (Tuesday and Friday) and add some users actually traveled by taxitaxi
◦ Regroup users living in nursing home and users with cognitive Regroup users living in nursing home and users with cognitive problems on same days, if possibleproblems on same days, if possible
◦ no transfer of users actually traveled by bus to taxisno transfer of users actually traveled by bus to taxis◦ transfer users actually traveled by taxi to buses if it induces a transfer users actually traveled by taxi to buses if it induces a
cost reductioncost reduction
INFORMS Healthcare 2011INFORMS Healthcare 20119
Sector 1Sector 1◦ 7% reduction of total traveling time 7% reduction of total traveling time ◦ 11% reduction of total distance11% reduction of total distance◦ 3 taxi users added to routes3 taxi users added to routes◦ reduction of $93 per week on cost: considering 45 operating weeks reduction of $93 per week on cost: considering 45 operating weeks
this a reduction of $this a reduction of $4,100 per year4,100 per year
Sector 2Sector 2◦ 1.3% increase of total traveling time 1.3% increase of total traveling time ◦ 8% increase of total distance8% increase of total distance◦ 17 taxi users added to routes17 taxi users added to routes◦ reduction of $85 per week on cost: considering 45 operating weeks reduction of $85 per week on cost: considering 45 operating weeks
this a reduction of $this a reduction of $3,100 per year3,100 per year
INFORMS Healthcare 2011INFORMS Healthcare 201110
Consider m antsConsider m antsStart with an initial solution for each ant Start with an initial solution for each ant kk◦ contruct a route with all userscontruct a route with all users◦ ijij = 1 for all edges ( = 1 for all edges (i,ji,j) and all ) and all kk
Update pheromone trailsUpdate pheromone trailsFind new routesFind new routesEnd if termination conditions satisfiedEnd if termination conditions satisfied
INFORMS Healthcare 2011INFORMS Healthcare 201111
INFORMS Healthcare 2011INFORMS Healthcare 201112
ParametersParameters
Probability of an ant Probability of an ant kk selecting an edge ( selecting an edge (i,ji,j) )
INFORMS Healthcare 2011INFORMS Healthcare 201113
( , )ij i j amount of pheromone on 1
ij ijij
d i jd
where is the distance between and
ki kΝ feasible neighborhood of ant
ki
ij ijkij
il ill N
INFORMS Healthcare 2011INFORMS Healthcare 201114
1
0 1
( 1) (1 ) ( ) ( ) ,
where is the trail evaporation rate and is the number of ants
at iteration m
kij ij ij
k
m
t t t t
1( , )
( )
0
k kij
i j kL t
if arc is used by ant
otherwise
( ) ( , )
( )
amount of pheromone ant deposits on arc
length of the th ant's tour
kij
k
t k i j
L t k
Values of parametersValues of parameters
INFORMS Healthcare 2011INFORMS Healthcare 201115
1
1
0,1
1 for all edgesinitial
Offers the possibility to:Offers the possibility to: optimize routes when locations of users are knownoptimize routes when locations of users are known visualize routesvisualize routes generate detailed instructions to follow the routegenerate detailed instructions to follow the route keep routes in a data base for a later usekeep routes in a data base for a later use modify routes if new users are added or removed or if modify routes if new users are added or removed or if
the type of mobility is modifiedthe type of mobility is modified interact with GoogleMapinteract with GoogleMap
INFORMS Healthcare 2011INFORMS Healthcare 201116
INFORMS Healthcare 2011INFORMS Healthcare 201117
InterfaceInterface
INFORMS Healthcare 201118
InterfaceInterface
INFORMS Healthcare 2011INFORMS Healthcare 201119
InterfaceInterface
Asymmetric TSP problem with additional constraintsAsymmetric TSP problem with additional constraints Use of heuristics to solve the problem: greedy Use of heuristics to solve the problem: greedy
algorithm and ant colony approachalgorithm and ant colony approach Significant improvement of actual solutions with a Significant improvement of actual solutions with a
simple greedy heuristicsimple greedy heuristic User friendly tool to help administrators generate User friendly tool to help administrators generate
routes on an everyday basisroutes on an everyday basis
INFORMS Healthcare 2011INFORMS Healthcare 201120