Review of ‘Agent-Based Modeling and Simulation’, OR Essentials

Book review to be published in The Journal of Artificial Societies and Social Simulation (JASSS)

Taylor, Simon J. E. (Ed.) (2014) Agent-Based Modeling and Simulation, OR Society and Palgrave Macmillan: Basingstoke

 

Agent-Based Modeling and Simulation is the first in the Operational Research Society’s OR Essentials series.   OR Essentials brings together multidisicpinary research from the management, decision, and computer sciences.  This edition within the series is edited by Simon Taylor, who is co-founder of the Journal of Simulation, also published by the OR Society.

The edited book is divided into 14 chapters, and the bulk of its contents convers the application of agent based modelling (ABM) to specific problem domains.  The book adds to this by introducing agent-based modelling as a technique, and setting it in context with other simulation approaches.  A very helpful chapter by Macal and North of Argonne National Laboratory offers a tutorial on what agent-based modelling is, focusing on the autonomy and interconnectedness of agents, and showing how agent-based models should be built.  The book ends with thoughtful chapters on a testing framework for ABM (Gürcan,  Dikenelli, Bernon), and a comparison with discrete-event simulation by Brailsford, elegantly closing the package opened by Taylor’s introduction comparing ABM with system dynamics and discrete event simulation within the context of modelling and simulation more generally.

The academic rigour of the book is confirmed by each article being reprinted from published articles from the Journal of Simulation.  The book brings together several excellent examples of agent-based modelling, together with a very clear understanding of how ABM fits in with more traditional simulation techniques such as DES (Discrete Event Simulation) and SD (System Dynamics) –  both Talyor and Brailsford show how and when ABM should be used.  Macal and North offer a very useful tutorial for understanding the building blocks of an ABM simulation, while Heath and Hill show ABM’s evolution from cellular automata and complexity science through complex adaptive systems.

Domain specific chapters cover applications in the management of hospital-acquired infection (Meng, Davies, Hardy, and Hawkey); product diffusion of a novel biomass fuel (Günther, Stummer, Wakolbinger, Wildpaner); urban evacuation (Chen, Zhan); people management (Siebers, Aickelin, Celia, Clegg); pharmaceutical supply chains (Jetly, Rossetti, Handfield); workflow scheduling (Merdan, Moser, Sunindyo, Biffl, Vrba); credit risk (Jonsson); and historical infantry tactics (Rubio-Campillo, Cela, Cardona).

Agent-Based Modeling and Simulation offers a very useful collection of applications of ABM, and showcases how ABM can be successfully incorporated in to mainstream, published research.  The contributions to the book are diverse, and from internationally regarded scholars.  It is also useful to see the diverse ways that agent-based modelling research is presented, from some papers that show code, some that show running models, and some that do not show the model or code but instead describe results.

The glue that binds the book is methodological.  Seeing how ABM has been used in diverse application areas is important, given the trans-disciplinary nature of the approach.  It is an excellent introduction into agent-based modelling within a wide range of business and operations applications, and should be read by scholars and practitioners alike.

Building the Multiplex: An Agent-Based Model of Formal and Informal Network Relations

EURO Conference 2016 PoznanThis presentation from the EURO 2016 conference in Poznan, Poland, and from the GDN conference in Bellingham, WA, USA, joint work with Leroy White of Warwick Business School, shows how combining formal and informal organizational networks enables decisions to flow more freely around organizations, but at a cost, leading to an optimal size of informal organizational networks.  If organizations can control these, this leads to implications for optimal information flows in companies.

What proportion of an airline ticket is made up of the cost of the aeroplane?

aiga-40_departingflightsAircraft aren’t cheap.  Neither are airline tickets.  But how much of that airline ticket is made up of the cost of the aeroplane?

If we assume a relatively efficient modern airliner, say a 777, a 30-year lifetime, 3500 hours per year, and an average speed of 500mph, that produces a  total distance of 52,500,000 miles.  Which is quite a lot.

If you were to knock on Boeing’s door, they could sell you one for $320 million.  Volume discounts are, I am told, available.

So, assuming straight line depreciation, along with many, many other assumptions, that’s $6.10 per mile.  At 350 or so passengers, that is around

2 cents per mile

Or, for a 3000 mile (transcontinental or transoceanic) flight, a total of sixty dollars.  Which is perhaps more than I expected.

Stacking Shelves the Amazon way

Amazon2Amazon FC are playing in the Euros (the UEFA football championship).  Or at least that’s what could be inferred from my name badge.

In fact, Amazon FC is one of the Amazon fulfilment centres, located in the Polish town of Poznań, location of the 28th European conference on Operational Research.

The fulfilment centre is huge – with a million separate items stocked, and up to a million items being processed every day – and is just one of a network of existing sites around Europe and around the world, the locations of which are themselves optimized to minimize cost.

AmazonMapOne of the many interesting facts about the tour was the way that Amazon store stock on their shelves.  Like any other business, they want to minimize fixed costs.  One way they can do this is to maximize the density of items stored on their shelves.  Unlike in my Mini factory visit (which will be the subject of a later post), Amazon does not run a just-in-time stock system.  They are happy(ish) to hold stock on the basis that their customer will have it quickly and will not have to wait for it to be backordered.  Amazon was founded on the basis of being able to supply items that only a few people will want – the so-called long tail – itself the subject of a book available on, where else, Amazon… the upshot of which is that if you rank the most popular to the least popular items sold on one axis, and take a logarithm of the number of these items sold on the other, it will make a nice straight line (a Zipf distribution for those interested).  So, Amazon will hold on to some items for years on the basis that someone, somewhere, sometime, will want to buy it.

AmazonFulfilment

So, Amazon needs to store these millions of items.  While most things are controlled and optimized by computer, they leave it down to human intuition as to where to store items – albeit guided by optimization algorithms.

AmazonVideo

Instead of giving a specific destination for each stock item (basically a very large grid reference), they give their employees a general area into where to store items.  The idea behind this is that when you first fill a location with stock, the space-packing density will be high as items fit next to each other.  But as items are removed and sold, spaces will be created meaning that you are effectively paying to store air.  So Amazon allows its employees to use these spaces to store more items – even if they are unrelated to each other.  Of course, the system keeps track of stock locations, but by doing things this way, the efficiency of the operations is improved, and less space is required for storage.

There is a similar example of adaptive organization used by Southwest airlines in my article The Complexity of the Corporation.

Agent-Based Modeling and Behavioral Operational Research

borBehavioral Operational Research: Theory, Methodology and Practice (Martin Kunc, Jonathan Malpass, Leroy Wright, Eds.) was published by Palgrave Macmillan in September 2016.  My chapter on Agent-Based Modeling and Behavioral Operational Research shows the great potential of using agent-based simulation within BOR, showing how example models can be applied to the field.  More details of the chapter can be found on the Palgrave Macmillan site here (DOI:10.1057/978-1-137-53551-1_7).

xbehavioral-operational-research.jpg.pagespeed.ic.mMtEvplVr7Please click here for the chapter in pdf format.

Updated September 2016 with full text.