You are on page 1of 3

C D

Vehicle Routing and Scheduling


5
T U T O R I A L*

DISCUSSION QUESTIONS a tour must be constructed for all vehicles in the fleet and
1. (a) Routing problems concern determining the order in capacity for each vehicle is not unlimited, thus capacity
which a single vehicle or several vehicles are to visit must be considered.
delivery or pick-up points. The order in which these 3. For the mail-delivery activity, the problem is a CPP.
points (called nodes) are visited is called a tour, the However, it also has aspects of a routing and scheduling problem
starting point of which is called the depot node. The in that the postman must find the best route for the bulk mail
objective is generally to minimize travel time or cost. pickups while considering the schedule for these pickups. The
Types include the traveling salesman problem, the pickup may be described as having a one-sided time window.
multiple traveling salesman, and the vehicle routing 4. (a) Deadhead time is nonproductive travel time incurred in a
problem. scheduling problem, i.e., the time to travel to a job.
(b) Scheduling problems deal with the order in which tasks (b) A depot node is the point of origin for a tour. It may be
are to be done. Generally these tasks have time arbitrarily assigned in most routing problems.
constraints associated with them, meaning that the start (c) An undirected arc means that a vehicle can travel in either
and/or ending times are specified as part of the problem. direction on the arc in the same time or at the same cost.
The nodes, to use the general terminology, are the tasks.
5. (a) A feasible VRP tour means that the sum of the demands
The order in which the tasks are done is the schedule. The
from the nodes assigned for the vehicle does not exceed
objectives may be to minimize cost or idle time.
the capacity of the vehicle and each node in a vehicle’s
(c) Combined routing and scheduling problems seek to
tour is visited only once.
route vehicles through pick up and delivery points while
(b) A feasible tour for a TSP means that each node in the tour
taking task-time constraints into consideration. Thus these
is visited only once; the tour begins and ends at the depot
problems seek to find not only the order in which nodes
node.
are to be visited but also the time given that some nodes
(c) A two-sided time window in a scheduling problem means
can be visited only within a specified time period.
that the vehicle or server can arrive at a depot (task)
2. (a) The classic traveling salesman problem (TSP) involves within a specified interval. For example, deliveries to a
routing of a single vehicle though a large number of nodes Washington, D.C., restaurant may be made only between
at which a pickup or delivery may occur and where the 11 A.M. and 1 P.M.
vehicle does not have a limitation on capacity. No time (d) A node precedence relationship means that one node must
constraints are involved. An example could be the route a be visited before another node. This situation may occur
Girl Scout takes in delivering cookie orders, although it is when a pick-up must occur before the delivery of the
doubtful a Girl Scout would use this approach. same product.
(b) In contrast to the TSP, in the Chinese postman problem
6. Both the nearest neighbor procedure (NNP) and Clark &
(CPP), demand (pickup or delivery) occurs in the arcs of
Wright (C&W) are heuristics for constructing tours for a TSP,
the tour, rather than at the nodes. An example is refuge
MTSP or VRP. The NNP is short-sighted in that it builds upon
pick-up or snow removal. As with the TSP the tour begins
partial tours by selecting the node which is closest to the last node
and ends at a depot node. In this type of problem, the
added. The C&W heuristic coincides the least “cost” order in
nodes (in the TSP sense) are so close together they cannot
which to “hook-up” nodes such that a “minimum” cost time tour
be considered discretely, thus it is easier to deal with
is constructed. Thus, the C&W heuristic evaluates the best order
demand on the arcs rather than at the nodes.
in which to attach nodes prior to constructing a tour, whereas the
(c) In a vehicle routing problem (VRP), there are several
NNP only looks ahead to a single node addition.
vehicles which travel from a single depot. The vehicles
have varying capacity, and this affects the route in that 7. Whenever the time or the cost required to move between a  b
a single vehicle can visit a limited number of nodes. or b  a, is different, the matrix is asymmetric.
As with the TSP, demand occurs at the nodes. In contrast, Such a situation may arise if there is construction in one
direction of travel and not on the other, or if there is not a parallel
*Source: C. Haksever, B. Render, R. Russell, R. Murdick, Service Manage- return path.
ment and Operations, 2nd ed. (Upper Saddle River NJ: Prentice Hall, 2000).

308
CD TUTORIAL 5 V E H I C L E R O U T I N G A N D S C H E D U L I N G 309

 8. Objectives that might be used to evaluate routes and schedules: If we continue with the 7–1 possibility, the hook-ups are:
(a) School buses: (1) Minimizing travel time to school; (4) 1–2
(2) Minimizing cost. (5) 2–3
(b) Furniture delivery trucks: (1) Minimizing travel time; (6) 3–6
(2) Minimizing dead-head (7) 6–8
time. (8) 8–4
(c) Ambulances: (1) Minimizing response time; The first possible tour is therefore:
(2) Maximizing the level of service; 4–5–7–1–2–3–6–8–4; 39.1 miles
(3) Minimizing time to a hospital.
Alternatively, we could have selected 7–6 at branch 3.
 9. (a) Mass Transit System: (1) Breakdowns; The resulting tour would then be:
(2) Construction;
4–5–7–6–1–2–3–8–4; 39.3 miles
(3) Rush-hour demand.
(b) A National Trucking Fleet: (1) Changes in demand; Note that the NNP cannot guarantee the same total tour
(2) New markets; length for all tours when multiple solutions are present.
(3) Maintenance of the fleet. In this case we must “cost out” all of the alternatives
(c) Snow Plows: (1) Variable demand due to a harsh winter; and select the least cost tour from the set.
(2) Vehicle maintenance; T5.5 The savings matrix generated from the C&W heuristic is
(3) Phasing road traffic with plowing (in miles):
(e.g., in a city in rush hour).
3 4 5 6 7 8
10. The savings in the C&W heuristic refers to the cost (in terms
of distance or dollars) of not connecting two nodes into a partial 2 3.9 2.6 1.4 1.3 0.8 0.0
tour or route. 3 — 6.6 4.7 5.3 2.7 0.2
4 — 6.0 6.2 4.0 0.9
5 — 9.9 6.6 2.6
END-OF-TUTORIAL PROBLEMS 6 — 8.5 3.6
T5.1 Assume that node 1 is the depot node. 7 — 3.2
(a) S23 = C12 + C13 – C23 = 10 + 14 – 5 = 19 miles
(b) S34 = C13 + C14 – C34 = 14 + 12 – 6 = 20 miles For example, the savings S23 = 22 + 5.8 – 4.1 = 3.9 miles.
(c) S25 = C12 + C15 – C25 = 10 + 16 – 18 = 8 miles Also, Sij = Sji for all i and j.
First, rank the savings from highest to lowest. Move
T5.2 The cost of inserting node 2 between 1 and 3:
down the list, linking nodes which have not been linked at
(a) I13 = C12 + C32 – C13 = 5 + 6 – 7 = 12
a previous step.
The cost of inserting between 3 and 5:
Steps:
(b) I35 = C32 + C52 – C35 = 7 + 8 – 8 = 7
(a) Link 5 and 6; (5–6)
The cost of inserting between 5 and 1:
(b) Add node 7; (5–6–7)
(c) I51 = C52 + C12 – C51 = 8 + 5 – 9 = 4
(c) Note that the 6–8 link is next in the list, however, to
Node 2 should be inserted between nodes 1 and 5. link these nodes we would have to break the subtour
T5.3 As each vehicle can be assigned (or not assigned) to each 5–6–7. Skip this link and move to the next feasible link.
of the 20 nodes, the number of possible alternatives is 220. (d) Link 3 and 4; (3–4)
T5.4 (a) Nearest neighbor heuristic solution with depot node = 1. (e) Link the Step d subtour with the other subtour (5–6–7)
Hook-up nodes: to yield 3–4–5–6–7
(1) 1–2 (f ) The 10th ranked link is 2–3; the new subtour is
(2) 2–4 2–3–4–5–6–7
(3) 4–5 (g) Link 8 and 7; 2–3–4–5–6–7–8
(4) 5–7 (h) The tour begins and ends at the depot node, thus the
(5) 7–6 solution is: 1–2–3–4–5–6–7–8–1 with a total tour
(6) 6–8 length of 27.3 miles. Compare this with the NNP
(7) 8–3 solution in Problem T5.4 which was 33.9 miles.
(8) 3–1 T5.6 (a) The NNP for vehicle 1 yields 2 tours: 1–2–4–3–5–1 at
Tour: 1–2–4–5–7–6–8–3–1: 37.9 miles a cost of $134 and 1–3–4–2–5–1 at a cost of $152.
(b) Nearest neighbor heuristic solution with depot node = 4. For vehicle 2, three tours are possible:
Begin with node 4 and hook-up: 1–6–9–8–7–10–1 ($204); 1–6–10–9–7–8–1 ($209);
(1) 4–5 and 1–6–10–9–8–7–1 ($188). When multiple solutions
(2) 5–7 are present, the lowest cost solution should be selected,
(3) 7–1 or 7–6 as each is 3.6 miles, thus there are ($134 and $188, for vehicles 1 and 2, respectively).
2 alternatives.
310 CD TUTORIAL 5 V E H I C L E R O U T I N G A N D S C H E D U L I N G

T5.7 Using the tours generated by the C&W heuristic,


vehicle 1 has a passenger load of 30;
vehicle 2 a passenger load of 45.
The total tour cost for both vehicles is $320.
As we are trying to improve the overall routing, we
attempt to switch a node from the longest to the shortest
tour (Tour 2 to Tour 1). Only nodes 6, 7 and 10 are
candidates as Vehicle 1 only has five seats remaining.
Consider switching node 6 into tour 1. The costs of
insertion at each link in tour 1 are:
I12 = $28
I24 = $65
I43 = $70
I35 = $37
I51 = $14
The minimum cost is achieved by inserting node 6 between
nodes 5 and 1; the tour would thus be 1–2–4–3–5–6–1 with
a cost of $148. Tour 2’s cost would be reduced to $177.
The overall cost is $325 which is greater than the initial
configuration so the switch should not be made. We must
also consider nodes 7 and 10 to determine if a switch can
improve the tour, by reducing the overall cost.
T5.8 The cost matrix is:

1 2 3 4 5 6 7 8
1 — 107.26 118.11 113.20 116.50 123.50 111.88 111.88
2 — 113.53 111.88 118.10 125.30 116.50 118.10
3 — 110.56 118.70 120.50 119.90 124.90
4 — 109.90 119.10 111.88 117.90
5 — 111.88 106.60 118.50
6 — 111.88 123.50
7 — 113.20
8 —
T5.
9 Vehicle 1 will be scheduled for tasks 1–3–6–8
Vehicle 2: 2–5–9.
Vehicle 3: 4–7–10.
Note that there are several solutions to this problem. The
above solution attempts to balance the need of all vehicles.

CASE STUDY
ROUTING AND SCHEDULING
OF PHLEBOTOMISTS 4. If you were to view this as a cluster first, route second
1. What characteristics of routing and scheduling are situation, based on what criteria would you form clusters?
exhibited in this problem? The easiest method would be clustering by proximity. A
Routing characteristics and one-sided time windows. second cluster might be severity of the cases or difficulty in
gathering specimens.
2. What type of data would you need to collect in order to
most effectively schedule technicians? 5. Suggest how you would solve the problem if the timed
The time it takes to collect a sample or the time spent on a specimens and routing pickups were considered separately.
floor, and travel times from one area to another (eg., the Routine pickups could be solved as a multiple TSP; the
travel time from Pediatrics to ICU). timed-specimen pickups would then be a straight scheduling
problem.
3. Does a deadhead time exist in this situation? If so, where?
Deadhead time exists in the scheduling aspects of the
problem. It is the time to walk from one collection site to
another.

You might also like