Impact

ISSN: 2058-802X (Print) 2058-8038 (Online) Journal homepage: www.tandfonline.com/journals/timp20

DISCOVERING HEURISTICS

JONATHAN THOMPSON

To cite this article: JONATHAN THOMPSON (2020) DISCOVERING HEURISTICS, Impact, 2020:1,
25-29, DOI: 10.1080/2058802X.2020.1739866
To link to this article: https://doi.org/10.1080/2058802X.2020.1739866

Published online: 03 May 2020.

Submit your article to this journal

Article views: 736

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=timp20
D I S COV E R I N G H E U R I S T I C S
JONATHAN THOMPSON IN A WORLD OF LIMITED WHAT ARE HEURISTICS
RESOURCES, there is growing AND WHEN SHOULD
pressure on businesses, organisations THEY BE USED?
and individuals to use these resources The word heuristic, from the Greek
effectively and prudently. This leads word heuriskein meaning find or
to many optimisation-type questions discover, refers to approaches to
– for example how should we route problem solving that do not guarantee
our delivery vehicles to minimise finding an optimal solution, but
CO2 emissions? How should we produce solutions of sufficient quality
allocate our nurses in order to meet to meet the needs of the problem
the medical needs of the ward? solver. In electing to use a heuristic,
How should we travel to a given the problem solver is accepting a
destination in an environmentally corresponding reduction in solution
friendly, yet timely manner? Such quality. However, there are many
decisions are being made on a scenarios in which exact solution
daily basis and the quality of these methods are not appropriate. These
decisions has a huge impact on the situations include:
usage of scarce resources. In academia,
a considerable amount of research • When no efficient solution meth-
focuses on making the optimal ods are known to exist. For the set
decision using methods such as linear of problems for which no efficient
programming, integer programming solution methods are known, even
and dynamic programming. However, powerful software for finding exact
for many business situations, time is solutions will struggle to solve to
an extremely scare commodity and optimality in a reasonable amount of
there is a need to produce solutions time other than for relatively small
in a timely fashion. Therefore, a problems. Many real-life problems
huge number of these decisions are fall into this category.
made using heuristic methods. • When time is limited. Even if
efficient methods are known, it may
be that the time required to find an
The word heuristic, from optimal solution is too long and the
the Greek word heuriskein problem solver requires a solution in
meaning find or discover, a shorter amount of time.
refers to approaches to • When the problem solver is willing
problem solving that do not to accept a decrease in solution
guarantee finding an optimal quality. For some problems, the
solution, but produce value of the objective function is
solutions of sufficient crucial and finding the optimal
quality to meet the needs of solution is absolutely vital. In some
the problem solver other contexts finding the precise

IMPACT © 2020 THE AUTHOR 25

 optimal solution is less important
and a reduction in solution quality
is acceptable.
• When the problem contains uncer-
tainty. In many practical situations,
the precise problem definition is
uncertain or includes uncertain
information. There is little value in
using considerable resources to find
an optimal solution to a vaguely
defined problem instead of using a
heuristic method.
• When the problem lacks infor-

© Kursat Unsal/Shutterstock
mation or includes contradictory
data.

In these situations, a heuristic

approach can be considered as the
solution method as exact methods
are not appropriate. In designing a FIGURE 1 MAP OF LONDON ATTRACTIONS
heuristic approach, it is important
that the problem is fully defined
and understood. Similar problems AN EXAMPLE– which no efficient solution methods
can then be identified and literature THE TRAVELLING are known, so heuristic methods are
can help to pinpoint attributes of SALESPERSON PROBLEM typically used in practice. Consider
heuristics that work well on related Consider a London tourist who the distance matrix in Table 1
problems. wishes to leave their hotel near Hyde showing the distance between the
Different types of heuristics exist Park and visit six famous landmarks six landmarks and the hotel. The
and one challenge is knowing which in the shortest distance before spatial arrangement of the hotel and
type of heuristic to use in a particular returning to their hotel. This is an landmarks can be seen in Figure 1.
situation. Heuristics may be classified example of the travelling salesperson The optimal route is Hotel -> London
as constructive, improving and meta- problem (TSP) which is an extremely Zoo -> Nelson’s Column -> Tower
heuristics. well researched problem and for Bridge -> Big Ben -> Buckingham

TABLE 1 DISTANCE MATRIX CALCULATED USING GOOGLE MAPS

BIG BUCKINGHAM CONCERT LONDON NELSON’S TOWER
HOTEL
BEN PALACE HALL ZOO COLUMN BRIDGE
Hotel 0 3.7 2.4 2.4 3.9 3.5 8.9
Big Ben 3.7 0 1.3 3.9 5.6 0.8 4.5
Buckingham
2.4 1.3 0 2.7 5.6 1.1 6.4
Palace
Concert Hall 2.4 3.9 2.7 0 6.1 3.9 9.2
London Zoo 3.9 5.6 5.6 6.1 0 4.8 9.0
Nelson’s
3.5 0.8 1.1 3.9 4.8 0 4.8
Column
Tower Bridge 8.9 4.5 6.4 9.2 9.0 4.8 0

26 IMPACT | SPRING 2020

Palace -> Concert Hall -> Hotel, consider slightly cleverer heuristics in the order given, which leads to
which has a distance of 24.4km. may give significantly improved the solution Hotel -> Big Ben ->
A constructive heuristic results. A further issue for heuristics Buckingham Palace -> Concert Hall
generates a solution from scratch is that the problem solver does not -> London Zoo -> Nelson’s Column
and repeatedly adds to the solution know whether the resultant solution -> Tower Bridge -> Hotel which has a
until it is complete. The simplest is close to optimal, or of very poor distance of 31.4. Our neighbourhood
example of a constructive heuristic quality. Lower (or upper) bounds may is defined as the set of solutions
for the TSP consists of going to the be used to give some context to the created by removing two edges from
nearest unvisited landmark until all results. the current solution and forming
have been visited, at which point An improvement heuristic takes a new feasible solution. In all cases
the salesperson returns to the hotel. some starting solution and gradually there is only one way of doing this.
In this example, the resulting route improves it. A neighbourhood An example is shown in Figure 2
would be Hotel -> Buckingham relationship defines those solutions where the links between Concert Hall
Palace -> Nelson’s Column -> Big that can be reached in one step and London Zoo, and Tower Bridge
Ben -> Concert Hall -> London Zoo from the current solution. A simple and Hotel have been identified for
-> Tower Bridge -> Hotel, which has improvement heuristic, also known as removal. To form a new valid route,
a distance of 32.2km, significantly local search, will accept a neighbouring the new links must be Concert Hall
further than the optimal solution. solution as the current solution if its to London Bridge and London Zoo
This approach is also called greedy solution quality is superior. to Hotel. The links that are removed
because it makes locally optimal are of distance 15 whereas the new
decisions at each step, but may incur Different types of heuristics links are of length 13.1. As the
large costs towards the end of the exist and one challenge neighbouring solution has lower cost
process. is knowing which type of than the current solution, the move is
However, other constructive heuristic to use in a particular accepted and the new tour is formed
heuristics can be used. Another situation which has a distance of 29.5. Note
simple heuristic involves choosing that the links between London Zoo
the cheapest link in turn and adding In the above example, we and Nelson’s Column, and Nelson’s
it to the tour as long as this does start from a simple starting point Column and Tower Bridge, are
not create subtours. We start by corresponding to visiting the cities now being traversed in the opposite
selecting Nelson’s Column -> Big Ben
(distance 0.8) and add Buckingham
Palace -> Nelson’s Column (1.1).
The next shortest link is between
Big Ben and Buckingham Palace
(1.3) but including this link would
result in a subtour so is ignored.
When completed, this heuristic
gives a solution of Hotel -> Concert
Hall -> London Zoo -> Tower
Bridge -> Big Ben -> Nelson’s
Column -> Buckingham Palace ->
Hotel, which has length 26.3km,
which although not optimal, is a
significant improvement over the
previous heuristic. This illustrates
that different heuristics can give FIGURE 2 ILLUSTRATING AN IMPROVEMENT HEURISTIC

very different solution quality and The diagram on the left shows the initial route. The links in orange have been selected for
spending a small amount of time to removal. The diagram on the right shows the route after new links are added.

IMPACT | SPRING 2020 27

direction. Neighbouring solutions will
continue to be sampled until some
stopping criteria is satisfied. Other
neighbourhoods will provide different
solutions, for example removing 3
links before reforming the tour.
Improvement heuristics give
locally optimal solutions but once
the neighbourhood contains no

© Andrey_Popov/Shutterstock.com
improving solutions, the search is
trapped and cannot escape. Many
meta-heuristic methods have been
proposed to overcome this problem,
including simulated annealing, tabu
search and genetic algorithms. These
are well-researched methods that have
proved effective on many practical
problems. We now consider examples efficient method. Vali-Siar et al. (2018) with m items in another. Tabu search
of heuristics being used to solve classic compare a constructive heuristic to a has proved particularly effective for
optimisation problems – scheduling, genetic algorithm to schedule hospital solving a number of different routing
routing and packing. operations and found that both problems including problems with
improved on manual methods, with time-windows and dynamic travel
the constructive heuristic performing times.
HEURISTICS FOR particularly well in reducing overtime
SCHEDULING PROBLEMS by over 80%.
Many scheduling problems have Compared to the
been solved using heuristic methods. previous manual methods,
When tasks need to be sequenced, HEURISTICS FOR the company report
known as the job shop scheduling ROUTING PROBLEMS savings in labour usage,
problem, ordering jobs according Several construction heuristics for a reduction in product
to some priority rule such as vehicle routing problems have been damage and improved
minimising slack time or earliest due widely used in practice. The Clarke loading quality
date can prove effective. To produce and Wright savings heuristic starts
examination timetables, exams can with a set of n routes where n is the
be ordered in terms of difficulty number of customers and each route Delgado-Antequera et al. (2020)
to schedule, e.g. in order of most consists of a journey from the depot use a greedy algorithm hybridised
clashing exams, and then allocated to a customer and back again. The with a variable neighbourhood
to the time period that ensures a heuristic then merges two routes into search method to determine routes
feasible timetable while spreading a single route as long as it causes a for waste collection vehicles in
exams out for students. decrease in the overall distance. There Antequera, Spain. They identify
The scheduling problems to which are several cluster first, route second routes that reduce the distance
heuristic methods have been applied are heuristics that group the customers travelled by more than 50%.
wide-ranging. For instance, Karapetyan into feasible sets, and then solve Louati et al. (2019) use a heuristic
et al. (2015) schedule image acquisition a TSP to form the route for each to route waste collection vehicles
from a satellite using several heuristic cluster. Many improvement heuristics and find solutions that reduce the
methods which all outperformed have been proposed, using a variety operational time by 23% and distance
human methods. They identified of neighbourhood definitions such by 7.3% compared to manual
simulated annealing as the most as exchanging n items in one route methods.

28 IMPACT | SPRING 2020

HEURISTICS FOR CONCLUSIONS problems that can be solved exactly
PACKING PROBLEMS Heuristics are used in a large number but currently heuristics are the right
Packing problems involve packing of practical scenarios where run and only option in many practical
items into containers. Constructive time is limited and a compromise in situations.
heuristics such as first fit decreasing solution quality is acceptable. The right
and best fit decreasing have long been heuristic to use in a particular situation
used to produce solutions to packing depends on several factors including
problems. The items to be packed the available time and resources, the
are sorted from largest volume to importance of the solution quality
smallest. First fit places each item in and the certainty of the problem
turn into the first bin in which it fits definition. Therefore, for each problem
whereas best fit places the item in the the “correct” heuristic to use will differ
bin in which the smallest empty space from situation to situation.
is left. Improvement heuristics often When a reasonable amount
move items to different positions or of time is available, a promising
boxes and implementations of m ­ eta- research area is matheuristics which
heuristic methods are commonplace. combines heuristic and exact methods
to improve solution quality. This
includes such methods as using exact
Heuristics are used in a methods to define neighbourhoods
large number of practical in improvement heuristics, using Jonathan Thompson is a senior
scenarios where run time is heuristics to improve bounds in lecturer in the School of Mathematics
limited and a compromise in branch and bound and dividing the at Cardiff University. He has
solution quality is acceptable problem into smaller components implemented heuristics to solve
that can be solved exactly, and many real-life problems including
then merging the solutions. Work manpower planning problems,
A real-life example comes from in quantum computing may yet scheduling hospital operations and
Atlas Copco, an American company lead to improvements in the size of timetabling problems.
that supply parts to mining companies
around the world, shipping packages in
sea containers. These containers have
to be packed in a way that minimises FOR FURTHER READING
damage to items, satisfying constraints Delgado-Antequera, L., R. Caballero, J. Sanchez-Oro, J. M. Colmenar and
on weight limits, weight distribution, R. Marti (2020). Iterated greedy with variable neighbourhood search for a
stability, and ease of loading while multiobjective waste collection problem. Expert Systems with Applications
145: 113101.
fitting as many items as possible into
the container. Olsson et al. (2020) Karapetyan, D., S. Mitrovic-Minic, K.T. Malladi and A.P. Punnen (2015).
use a greedy construction heuristic Satellite downlink scheduling problem: A case study. Omega 53: 115–123.

to generate load plans and a genetic Louati, A., L.H. Son and H. Chabchoub (2019). Smart routing for municipal
algorithm to fine tune the objective waste collection: a heuristic approach. Journal of Ambient Intelligence and
Humanized Computing 10: 1865–1884.
function for the construction heuristic.
The method has been successfully Olsson, J., T. Larsson and N-H Quttineh (2020). Automating the planning of
container loading for Atlas Copco: Coping with real life stacking and stability
implemented and generates suitable
constraints. European Journal of Operational Research 280: 1018–1034.
plans in a timely fashion. Compared
to the previous manual methods, the Vali-Siar, M. M., S. Gholami and R. Ramezanian (2018). Multi-period and
multi-resource operating room scheduling under uncertainty: A case study.
company report savings in labour usage, Computers & Industrial Engineering 126: 549–568.
a reduction in product damage and
improved loading quality.

IMPACT | SPRING 2020 29