Dr Robin Purshouse
Reader in Decision Modelling and Optimisation
Dr Robin Purshouse
Department of Automatic Control and Systems Engineering
University of Sheffield
Tel: (+44) (0)114 222 5618
Fax: (+44) (0)114 222 5683
Robin Purshouse received the MEng degree in Control Systems Engineering in 1999 and a PhD in Control Systems in 2004 for his research on evolutionary many-objective optimisation under the supervision of Peter Fleming. Commercial experience includes Logica plc (1999-2000), PA Consulting Group (2003-2007) and Rolls-Royce plc (2007-2008). He returned to the University of Sheffield in 2008 - initially as a Research Fellow in the School of Health and Related Research and, since 2010, as a Lecturer in the Department of Automatic Control & Systems Engineering. Robin was a principal architect of the Sheffield Alcohol Policy Model, alongside Alan Brennan, Rachid Rafia, and Petra Meier. He was the holder of an ESRC Future Research Leaders grant award (October 2012 - September 2014) and, with Peter Fleming and Carlos Fonseca, was General Chair of the seventh international conference in the world-leading series on Evolutionary Multi-Criterion Optimization (EMO) held in Sheffield in March 2013.
ACS125 Systems Modelling and Simulation
Jointly led by George Panoutsos and Robin Purshouse, this module aims to provide students with an introductory systems perspective on both the science and art of mathematical modelling. Robin provides an introduction to discrete, stochastic and hybrid systems in the Spring Semester. The module motivates the need for mathematical modelling through a series of six major case studies, spread throughout the academic year. A diverse range of systems is considered, with the Spring Semester focusing on the pharmacoeconomics of HIV treatment options, the effect of environmental variability on an automotive suspension system, and the digital control of a hydraulic extrusion press.
ACS108 Laboratory and Professional Skills
Robin leads on the Employability Theme within this wider module on professional and laboratory skills taught by the Level 1 team at ACSE. Within the theme, through a series on engagements spread across the academic year, students are encouraged to reflect on the competencies needed to be professional systems engineers and active citizens. Our current industrial guest speakers on the module are Marion Carrabin from Transport for London and Luke Creasey from Iconsys.
Robin's research aims to help improve how we identify and choose between possible solutions to a problem, with a particular focus on the process of policy appraisal. There are a number of factors that make policy appraisal a challenging research area:
|Multiple trade-offs||When deciding on a best course of action, there is usually more than one performance criterion, or objective, to be considered. If these objectives are in conflict with each other, then we are faced with multiple ‘best’ solutions offering different trade-offs between the objectives. In decision analysis we often use methods that attempt to identify these trade-offs, but if there are more than three objectives to consider then some of the most popular methods are likely to fail at this task.|
|Multiple stakeholders||In order to identify a single best solution, we need to introduce subjective preferences for different trade-offs between objectives. Most decisions will affect a multitude of different stakeholders, so we need to use methods that can build a consensus or arbitrate between different sets of values. Some of the best practice methods for achieving this are quite unpopular with the stakeholders themselves. Open questions exist for government decisions: should the preferences be those of society? How do we measure these? How do we achieve consistency between different decisions?|
|Deep uncertainty||In decision analysis, we try to measure how well each possible solution performs against each objective. For engineering design problems, we can often build mathematical models that can predict quite precisely how an option will perform once implemented. However, when considering social policy, our current models of social processes are subject to much more uncertainty. We need to develop methods that can improve our quantitative understanding of social systems and also think more about how to integrate attitudes to risk, and the value of delaying a decision to reduce uncertainty, into trade-off analysis methods.|
|Cognitive challenge||In decision analysis, the aims are to provide support to the individuals tasked with making a decision and to help develop a shared understanding of the problem and its potential solutions. Communication of trade-offs is vital to these aims, and this task starts to become challenging as the number of objectives rises beyond three. If our predictions are precise, then visualisation methods are available – but when our predictions are uncertain then these current methods are of limited value.|
|NIAAA||Calibrated Agent Simulations for Combined Analysis of Drinking Etiologies (CASCADE), with investigators A. Brennan, P. Meier, M. Strong, G. Valentine||2016-2021||$2,313,250|
|EPSRC / JLR||Optimization for robust design: Integrating model-based systems engineering with multi-criteria decision-making support in a distributed framework, with investigators P. Fleming, V. Kadirkamanathan, I. Giagkiozis||2014-2018||£1,074,420|
|EPSRC||Value-based Bayesian sequential analysis for fault monitoring, with investigator A. Mills||2013-2014||£18,099|
|ESRC||Complex systems modelling of alcohol consumption dynamics in the British population||2012-2014||£173,428|
|MRC||Alcohol policy modelling and evaluation, with investigators P. Meier, A. Booth, A. Brennan, R. Maheshwaran and M. Hernandez||2010-2013||£1,034,163|
|Purshouse RC, McAlister J. Multi-objective optimisation for social cost benefit analysis: an allegory. Lecture Notes in Computer Science 2013;7811:726-740.
|Sets out a practitioners' perspective on the multiple criteria approach to cost benefit analysis for major investment decisions. The major challenges are identified. The paper is part of an initiative to bridge the gap between the theory and practice of decision science.|
|Purshouse RC, Brennan A, Rafia R, Latimer NR, Archer RJ, Angus CR, Preston LR, Meier PS. Modelling the cost-effectiveness of alcohol screening and brief interventions in primary care in England. Alcohol and Alcoholism 2013;48(2):180-188.||Documents the first major extension of the Sheffield Alcohol Policy Model beyond pricing - to the provision of screening and brief intervention for patients with alcohol misuse problems. The modelling identifies that such programmes are likely to be cost-effective, and possibly cost-saving. This research was used to support guidance by the National Institute for Health and Care Excellence.|
|Wang R, Purshouse RC, Fleming PJ. Preference-inspired co-evolutionary algorithms for many-objective optimization. IEEE Transactions on Evolutionary Computation 2012; Early Access Article.||Presents a new class of methods for many-objective optimization that aim to overcome the problem of incomparability associated with Pareto-based methods by co-evolving a family of preferences alongside the candidate solutions. The empirical results show that a goal-based implementation of the method is highly competitive with the world-leading algorithms for a posteriori optimisation. This work builds on an EMO 2011 conference paper that demonstrated proof-of-concept results.|
|Purshouse RC, Meier PS, Brennan A, Taylor KB, Rafia R. Estimated effect of alcohol pricing policies on health and health economic outcomes in England: an epidemiological model. Lancet 2010;375(9723):1355-1364.||Appraises the impact of a minimum price for the retail sale of alcoholic beverages on population health in England. Across 47 health conditions, the paper estimates volumes for premature deaths, disease prevalence and hospitalisations. The results have been used by public health advocates to make the case for minimum pricing, which was included in the UK Government's 2012 Alcohol Strategy.|
|Meier PS, Purshouse R, Brennan A. Assessing the impact of policy options for alcohol price regulation: the importance of modelling population heterogeneity. Addiction 2010;105(3):383-393.||Proposes the methodology of the Sheffield Alcohol Policy Model. This model, and its findings, have made a central contribution to the national and international debate on setting a minimum price for the retail sale of alcoholic beverages.|
|Purshouse RC, Fleming PJ. On the evolutionary optimization of many conflicting objectives. IEEE Transactions on Evolutionary Computation 2007;11(6):770-784.||Demonstrates empirically how the performance of multi-objective evolutionary algorithms can deteriorate as the number of objectives is increased. The work builds on a CEC 2003 conference paper that introduced the term 'many-objective' to the EMO research field.|
|Purshouse RC, Fleming PJ. Conflict, harmony, and independence: Relationships in evolutionary multi-criterion optimisation. Lecture Notes in Computer Science 2003;2632:16-30.||Establishes the agenda for research into decision problems with more than three objectives over the next decade. Defines three classes of relationship - conflict, harmony and independence - that form the basis for a critical re-examination of multi-objective evolutionary methods.|