Post on 15-Jan-2016
CAPS RoutePro
• Routing Environment.
• Solution Methods.
• Backhauls.
• Dispatcher Interface.
• Demonstration.
Routing Environment
• Load and unload times.
• Compatibility rules.
• Routing Properties.
• Routing Methods.
Load and Unload Times
• Fixed time per stop.– Minimum time to service a stop.
• Fixed load/unload time.– Independent of what and how much is being
loaded/unloaded.
• Variable load/unload time.– In addition to fixed load/unload time.– Based on amount being loaded/unloaded.– Can round values to reflect partial loads (time to
move a pallet is same whether full or half full)
Load and Unload Times
Compatibility (Mateability)
• Define compatibility/incompatibility for:
– Site and vehicle.
– Work zone and driver or vehicle.
– Vehicle and order or product.
– Vehicle and driver.
– Vehicle and tractor.
– Product and product.
Driver and Vehicle Compatibility
Vehicle and Product Compatibility
Route Properties
• Which rules to enforce.– Capacities.– Time windows.– Compatibilities.
• Travel Properties.– Straight line distance or road network.– Access from sites to road network.– Out-of-route travel limits.
Enforcements
Travel Properties
For multiple stops at same location.
Adjust straight line distance for circuity.
Access to Road Network
• Sites are not on the road network.• Use “access roads” to nearest intersection
(node).• Can specify speed and circuity.
Travel a-b direct or along blue access roads?
a b
Out-of-Route Travel
Consider new orders only where out-of-route travel is <200 miles.
Out-of-Route Travel
3rd
2nd
1st
last
New stop
Add the new stop if it is not too far out of route
ab
c
x
d
e
if a + b + d + e <= x + 200
More Route Properties
• Deadhead miles.– For long-haul travel.– Limit empty miles as:
• % of route distance.• Maximum number of empty miles. • Maximum number of empty hours.
• Order Priorities.– Farthest from depot.– Largest quantity.– Shortest time window.– Earliest time window.
Deadhead Travel
Limit empty miles to 15% of total distance.
Deadhead Travel
D2
D1 P3
Dashed lines indicate deadhead travel
P4
D4
P1,P2
P = PickupD = Delivery
D3 x
a
Route must satisfy: c+g <= 0.15(a+b+c+d+e+f+g)
cb
d
f
g
e
Order Priorities
Distance from depot,
Quantity, or
Time window
Routing Methods
• Grow routes.– Construct new routes.– Add orders to existing routes only.– Add orders to existing routes first; then
construct new routes.
• Resequence stops on a route.– Do not move stops between routes.
• Use predefined routes.– Provides a baseline for existing routes.
RoutePro Solution Techniques
• Nearest Neighbor.• Nearest Pass. • Pie Algorithm. • Boundary Algorithm.• Cluster Algorithm.• Savings Algorithm.• Tree Algorithm.
• Weighted Constraint Algorithm.• Add to Existing Algorithm.
Solution Techniques
• Consider high priority orders first.
• Use route limits to consider only orders that do not increase length and cost “too much”.– Limit on added miles.– Limit on added hours.– Limit on added cost.
Route Limits
Prevents adding orders far from route.
Solution Techniques
• Build routes one order at a time.
• Location of first order placed on a route is called the seed.
• Link connecting depot and seed is called the skeleton.– Many algorithms try to add orders close to the
skeleton.
• Most methods require user specified angles to define potential orders to add.
Nearest Neighbor
• Add nearest order in the specified neighborhood to route if feasible.– Highest priority order is seed.– Neighborhood specified by distance and angle.– Only look within specified distance.– Consider small angles first, then expand.
Seed
Nearest Neighbor
Limit candidate orders to within 50 miles of seed.
Nearest Neighbor
Nearest Pass
• Add order with least incremental cost to an existing route.
• Start new routes if no existing route can take a new order.– Highest priority order is seed.
• Consider orders only in region around depot with specified angle.
Nearest Pass
Limit candidate orders to within yellow region or within 1 mile of depot.
Pie Algorithm
• Add order with smallest angle from skeleton within a specified “pie”-shaped region centered at depot.– Highest priority order is seed.
• Consider orders only in pie-shaped region with specified angle - or orders close to the depot.
• Solution quality depends on angle specified.– What angle is best?
Pie Algorithm
Limit candidate orders to within yellow region or within 20 miles of depot.
Boundary Algorithm
• Multiple use of Pie Algorithm with a user specified set of angles to increase size of “slice”.
• Specify – Minimum angle.– Maximum angle.– Number of increments.
• Good to find best angle for Pie Algorithm.
• Good for sparse distributions of orders.
Boundary Algorithm
Repeats Pie Algorithm for larger and larger slices.
Cluster Algorithm
• Adds orders sequentially for sites close together.
• New orders must be within a user specified distance of the route and not too near the depot.
• Good if orders are clustered in space and time.
• Not too good if clustering is not apparent.
Savings Algorithm
• Add orders to routes based on greatest “savings”.
• Savings reflect cost to link two orders on a route.
• Consider orders only in region of specified angle around depot.
Savings Algorithm
Limit candidate orders to within yellow region or within 1 mile of depot.
Tree Algorithm
• Adds orders to routes sequentially to minimize a weighted sum of two distances– from order to seed location. – from order to depot.
• User specifies weight to control emphasis on distance to depot or to seed for route.
• Useful to minimize empty travel on first and last legs of route.
Tree Algorithm
Tree Algorithm
Seed
Order 1
Order 2
a
c
b
d
Add Order 1 to route since a+b < c+d
Should order 1 or 2 be added next?
Weighted Constraint Algorithm
• Adds orders to existing routes to minimize distance.– Does not build new routes.
• Does not require user specified angles.
• Useful for orders that are hard to route.
Add to Existing Algorithm
• Adds orders to existing routes based on “regret”.– Does not build new routes.
• Regret = difference between cost with order on best and second best route.
• Assigns order with highest regret to best route.
Route Improvement Algorithms
• After building initial routes.
• Resequence stops on a route (intra-route improvements).
• Manually move stops or add stops.
Backhauls
Compare profit from deadheading and carrying a backhaul.
Pickup
Delivery
depot
D D
Empty Return
Backhaul
Profitable Backhauler Algorithm
• Requires revenue for potential backhauls to be given.
• Considers adding potential backhauls to existing routes:– Compares net profit for backhaul (revenue - cost)
with cost to deadhead.– Inserts backhaul with largest net profit.
Dispatcher’s Process
• Import orders and validate.
• Select orders to route.
• Build routes.– Build new routes and edit routes.– Use one or more route methods.– Limited to methods allowed by Administrator.
• Analyze results.
• Export routes.
Demo: New England Acme Tools
• 1 Warehouse
• 425 Customers – Hardware stores. – Locksmith shops.
• 2 Vehicle Types– 48 ft. trailers (25)– 28 ft. trailers (15)
• 130 Orders
Zoomed In View
5 Hardware stores 4 Locksmiths
Nearest Neighbor vs. Nearest Pass
16 routes, $10,254 11 routes, $9,795
Pie Algorithm vs. Boundary Algorithm
16 routes, $9,587 13 routes, $9,122
Cluster Algorithm vs. Savings Alg.
21 routes, $8,650 - 108 orders 20 routes, $16,583
Tree Algorithm vs. Weighted Constraint
11 routes, $8,600 12 routes, $9,962
Route Summary
Cost and Distance
Time Usage
Excel Vehicle Utilization Chart
Vehicle Utilization Chart
0100020003000400050006000
DRLG1
DRLG10
DRLG11
DRLG12
DRLG13
DRLG14
DRLG15
DRSH1
DRSH10
DRSH11
DRSH12
Vehicle ID
Max
vs.
Use
d C
apac
ity
PEAKQTY1
MAXCAPACITY1
Excel Vehicle Use Chart
Vehicles Used In Time
0
2
4
6
8
Mar/1712:00
Mar/180:00
Mar/1812:00
Mar/190:00
Mar/1912:00
Mar/200:00
Mar/2012:00
Route time
Num
ber o
f veh
icle
s us
ed
RoutePro Summary
• Very powerful package.
• Handles wide range of routing problems and real-world complexities.
• Allows user (administrator) control over all aspects.
• Provides multiple solution algorithms.– Must select best algorithm for each application.