ASOR Seminar Schedule

ASOR's Thursday Afternoon Online Seminars run weekly at 4.30pm AEST or AEDT (4pm SA, 1.30pm or 2.30pm WA depending on east coast daylight saving, etc) using Zoom. They are open to members and non-members, and feature a 20-25 minute presentation plus Q&A afterwards. The Zoom meeting details are sent by email to our mailing list subscribers.

We record these sessions with the presenters' permission and post them on our vimeo account with links from this page.

2025 Series

13 March 2025
Elli Irannezhad (UNSW)
Freight transport agent based modelling

27 February 2025
Guvenc Dic (QUT)
(title tba)

13 February 2025
Hajar Sadegh Zadeh (U Melbourne)
A Preventive Scheduling Model for Managing Elective and Emergency Surgeries in a Stochastic Setting

Australian hospitals face growing pressure to deliver medical services efficiently. Operating rooms (ORs) are essential yet challenging resources to manage due to their complex interplay of objectives, constraints, and uncertainties. This study introduces a two-stage stochastic model for scheduling both elective and emergency surgeries, addressing the unpredictability of emergency arrivals and surgery durations. The model’s first stage involves making preliminary scheduling decisions based on predictions, while the second stage refines these decisions with real-time information. Scenarios are generated from synthetic data inspired by the operating theatre of a public hospital in Australia, capturing daily emergency arrivals and dynamically adjusting schedules to improve efficiency. This synthetic data, reflecting historical trends and operational characteristics of the healthcare system, highlights key operational challenges such as emergency cancellations and under-utilization of elective surgery blocks. The model incorporates flexible OR allocations shared between elective and emergency cases and introduces buffer times to better accommodate unexpected emergencies and optimize resource use. Using sample average approximation, the model ensures robust decision-making under uncertainty. Results show substantial improvements in performance metrics, with total costs reduced by approximately 18%, OR utilization rates rising from 78% to 84%, and a marked reduction in emergency surgery cancellations. This study demonstrates the potential of advanced stochastic models to enhance hospital operational efficiency and effectively address the complexities of OR scheduling

30 January 2025
Efat Fakhar (Rio Tinto)
Application of Advanced Analytics: From PhD Research to Industry (view video)

In my PhD research, I explored the application of optimisation techniques to solve complex problems in gas distribution networks. The process of transporting natural gas from fields to consumers involves various challenges across extraction, processing, and distribution stages, each of which can benefit from Operations Research methodologies. My research focused on the design and allocation problem in natural gas distribution networks, developing a Mixed-Integer Nonlinear Programming (MINLP) model. This model determines the optimal network layout, pipeline diameters, node pressures, and flow through the links, all aimed at meeting demand at minimal cost. The model incorporates constraints related to gas laws, pipeline diameters, pressure limits, and network structure, which is modeled as a tree. To solve this, I employed both an approximation method and a heuristic approach, with computational results demonstrating the model's effectiveness. Beyond my academic project, I will discuss the broader application of optimisation techniques in industry. By integrating optimisation with advanced analytics, we can improve decision-making and operational efficiency, offering significant value across various domains.

2024 Series

5 December 2024
Evan Shellshear (CEO of Ubidy, and adjunct at QUT and UQ)
What is stopping OR and data science projects from realising their potential? (view video)

The field of artificial intelligence, data science, and analytics is crippling itself. Exaggerated promises of unrealistic technologies, simplifications of complex projects, and marketing hype are leading to an erosion of trust in one of our most critical approaches to making decisions: data driven. This talk, based on the book Why Data Science Projects Fail, aims to fix this by countering the AI hype with a dose of realism and guidance. The book was written by two experts in the field, Doug Gray and Evan Shellshear, who firmly believe in the power of mathematics, computing, and analytics, but understand that if false expectations are set and practitioners and leaders don’t fully understand everything that really goes into data science projects, then a stunning 80% (or more) of analytics projects will continue to fail, costing enterprises and society hundreds of billions of dollars, and leading to non-experts abandoning one of the most important data-driven decision-making capabilities altogether. Join us for a session sharing the insights from the book on why data science projects fail to make sure your projects don’t belong to the 80% of failures.

Dr Evan Shellshear is the Managing Director and Group Chief Executive Officer of Ubidy, an innovative global recruitment marketplace leveraging AI to connect employers to specialist agencies. He has a Bachelor of Arts and a Bachelor of Science (single and double majors in mathematics) from The University of Queensland, a Diplom (equivalent of both a BSc and MSc) from the University of Bielefeld (Germany) and a PhD in Mathematical Economics (Game Theory) from the Institute of Mathematical Economics at the University of Bielefeld. Before his appointment at Ubidy, Evan was Head of Analytics and Co-Chief Executive Officer at Biarri – one of Australia’s leading OR consultancies. He has published or co-published four books on topics in artificial intelligence, business analytics, and data science. He holds Adjunct academic appointments at the Queensland University of Technology and at The University of Queensland.

21 November 2024
Gaurav Singh (Decisions360)
The Next-Generation Fast MIP Solvers Using Parallel and Distributed Computing (view video)

Branch-and-Bound (B&B) methods are fundamental to solving Mixed Integer Programming (MIP) problems but face significant scalability challenges in parallel environments, particularly due to communication overhead. This talk introduces a novel approach to develop a parallel MIP solver that leverages advanced techniques for efficient B&B tree management. By addressing these limitations, the proposed solution aims to achieve rapid convergence, offering transformative benefits across critical industries such as logistics, finance, and manufacturing. This work represents a step forward in enhancing the scalability and performance of MIP solvers in distributed computing environments.

Gaurav brings over two decades of senior executive experience, focusing on decision sciences, digital transformation and technology in the mining industry and R&D sector. He is currently Managing Director & Founder of Decisions 360, an innovative start-up that is enabling businesses to make improved decisions through Decision Sciences. The work presented here is with our US partners Optimal Solutions Inc.

7 November 2024
Mostafa Khatami (U. Wollongong) - ASOR RISING STAR AWARD RECIPIENT 2023
Substitution or emergency order? Towards blood supply chain stability (view video)

Blood type substitution serves as a valuable strategy to enhance resilience in the blood supply chain. By leveraging compatibility among specific blood types, it becomes possible to more effectively balance the fluctuations in supply (donation) and demand (transfusion). However, a critical and complex decision lies in choosing whether to use a compatible blood unit from the on-hand inventory (substitution) or to place an emergency order. Data from Australia show that current practices have led to a marked imbalance between supply and demand, particularly for the highly compatible O-negative blood units. This study presents a stochastic optimization model utilizing sample-average approximation (SAA) to support hospitals in making blood-type substitution decisions. We evaluate the SAA model against several alternative policies, with results indicating that the proposed model significantly enhances blood unit management and mitigates the supply-demand imbalance for O-negative units.

Mostafa is a Lecturer in Business Analytics at the University of Wollongong (UOW), School of Business. He has a background in Operations Research, and a PhD in Mathematical Science from the University of Technology Sydney (UTS). His area of research expertise spans the fields of Operations Research and Computer Science. He enjoys solving scheduling, facility location, and supply chain problems, mainly arising in healthcare and transportation. Mostafa's research works have appeared in several prestigious OR/MS journals, including European Journal of Operational Research, Computers & Operations Research, Journal of Scheduling, and Journal of the Operational Research Society.

9 May 2024
Fatemeh Jalalvand (CSIRO Data61)
Seeking higher performance in a multi-objective simulation-based optimization method using Q-learning algorithm and a frequent pattern extraction method (view video)

Recently, there is an increasing number of studies on employing machine learning (ML) techniques to enhance optimization algorithms. However, the application of ML techniques to improve multi-objective optimization algorithms which can solve many real-world problems is under-explored. This study contributes by developing QLFPEMMOSO, which integrates two ML techniques as a Q-learning algorithm and a frequent pattern extraction method into a multi-objective simulation-based optimization method, containing Non-dominated sorting genetic algorithm (NAGA)-III and a system dynamics (SD) simulation model. QLFPEM-MOSO addresses two purposes. First, Q-learning selects the most performance-enhancing crossover operator among a set of crossover operators based on the concept of adaptive operator selection. Second, the frequent pattern extraction method extracts frequent patterns from a set of good solutions, i.e., the elite set, and injects them into the population of NSGA-III. Frequent patterns carry good knowledge that can lead NSGA-III toward generating better solutions. Meanwhile, NSGA-III searches for the optimal solutions and the SD simulation model calculates the fitness of the optimal solutions. Although QLFPEM-MOSO is a general method applicable to various problems, we show its applicability on a defense case study. The numerical results on various instances illustrate that QLFPEM-MOSO significantly outperforms the multi-objective simulation-based optimization method, which lacks any ML technique, respecting solution quality and convergence behavior. However, QLFPEM-MOSO imposes higher computational time. In certain instances, QLFPEMMOSO averagely saves $3B workforce cost (first objective) and improves capability gap (second objective) by 170.6, but adds computational time compared to the multi-objective simulation-based optimization method

7 March 2024
Simon Dunstall and Simon Knapp (CSIRO Data61)
A simulation-optimisation system for assisting bushfire firefighting fleet decisions (view video)

This talk describes the development of a MIP-based optimiser for assigning aircraft, helicopters and land tankers to fight bushfires, and a simulation system into which this optimiser is placed, so as to answer fleet sizing, positioning and deployment policy questions. The optimisation problem is a somewhat difficult one to formulate and solve, because the problem is naturally stochastic and the success of a firefighting mission depends on both past and future actions. The simulation system addresses firefighting fleet logistics and bushfire evolution over time, and calls the optimiser on an event queue to simulate the actions of a human and/or algorithmic fleet operational manager

8 February 2024
Richard Watson (Ryan Watson Consulting P/L)
A Geospatial Data Analysis of Environmental Factors in Victorian Road Accidents (view video)

This talk describes a project to identify geospatial environmental factors contributing to road accidents in Victoria and predict road accident hotspots using machine learning.. The project aims were first presented at MODSIM 2021. Much work of this nature has been done in other countries, but as far as we know little in the Australian context. A number of open datasets relating to road accident events, and geospatial and demographic environmental factors such as road characteristics, population density, traffic volume, tree cover and the built environment were examined using the Python geospatial analytics and machine learning libraries. Geospatial statistical techniques used include Moran I, Getis-Ord. Kernel Density Estimation and Kriging.

2023 Series

14 December 2023
Tiria Andersen (James Cook University)
Vehicle routing with spatial exploration (view video)

Autonomous vehicles, capable of in situ information collection and computing, are creating new opportunities in disaster relief, conservation management, surveying, and last-mile delivery. For vehicle routing problems, autonomous vehicles are agents that can seize the initiative to spatially explore and detect any new points of interest (POIs). This problem can be considered a vehicle routing problem with profits, or an orienteering problem - but with a spatial dependency, where the information state is strongly dependent on the vehicle's actions. The question arises: How do we route to known POIs while also strategically exploring areas with high probability of POI presence?

26 October 2023
Nariman Mahdavi Mazdeh (CSIRO Energy, Newcastle)
Data-enabled Predictive Control (DeePC) of Buildings (view video)

Nariman will provide an overview of the state-of-the-art data-driven control technique, the main motivations, and the similarities and differences with model-based control techniques, such as Model Predictive Control (MPC). DeePC is a recently developed approach, inspired by behavioural system theory, that combines system identification, estimation, and control in a single optimisation problem, for which only historical input/output observations of the system is required. Nariman will then share some insights about the DeePC application in building control, plus some initial simulation results after using the Building Optimisation Testing Framework.

14 September 2023
Yunzhuang Shen
Solution prediction via machine learning for combinatorial optimization (view video)

31 August 2023
Sameh Tawfiq AlShihabi (U. Sharjah)
A Novel Core-Based Optimization Framework for Binary Integer Programs- the Multidemand Multidimensional Knapsack Problem as a Test Problem (view video)

17 August 2023
Kyle Harrison (ASOR Rising Star Award recipient 2022)
Surrogate-Assisted Analysis of the Parameter Configuration Landscape for Meta-heuristic Optimisation (view video)

Real-world optimisation problems are often too complex for exact solution methodologies to address in a reasonable amount of time. Meta-heuristic optimisers can provide high-quality solutions to challenging problems in a reasonable amount of time but are highly sensitive to the values assigned to their control parameters. In fact, tuning control parameter values of a meta-heuristic can be thought of as an optimisation problem in itself. Thus, a valid question to ask is: how can we efficiently optimise the optimiser? Towards this goal, this talk will discuss the usage of artificial neural networks (ANNs) as surrogate models to greatly reduce the computational burden associated with characterising the parameter configuration landscape (PCL) of meta-heuristic optimisers. The trained surrogate models allow for constant-time estimation of the fitness associated with multiple executions of a parameter configuration, thereby facilitating an efficient way to sample and evaluate a large volume of parameter configurations. Ultimately, this can be used to design more effective parameter control strategies, which can then be used to improve optimisation outcomes.

3 August 2023
Honglei Xu (Curtin University)
Two-stage Games Under Uncertainty: Mathematical Formulation and Computation (view video)

20 July
Asef Nazari (Deakin University)
Metaheuristics for Optimising and Balancing Assembly Lines (view video)

2022 Series

1 December 2022
Qian Wan (CSIRO Data61)
The agricultural spraying vehicle routing problem with splittable edge demands (view video)

The capacitated arc routing problem (CARP) is to find a set of least-cost routes for a fleet of identical vehicles of limited capacity that must service the demand of a subset of edges in a network. We present a splittable agricultural chemical sprayed vehicle routing problem and formulate it as a mixed integer linear program. The main difference is that our problem allows us to split the demand on a single demand edge amongst robotics sprayers. We use theoretical insights about the optimal solution structure to improve the formulation and provide two different formulations of the splittable capacitated arc routing problem (SCARP), a basic spray formulation and a high-edge demands formulation. The solution methods consisting of lazy constraints, symmetry elimination constraints, and a heuristic repair method. Computational experiments on a set of valuable data based on the properties of real-world orchards reveal that the proposed methods can solve the SCARP with different properties. We also report computational results on classical benchmark sets from previous CARP literature. The tested results indicated that the SCARP model could provide cheaper solutions in some instances when compared with the CARP literature. Besides, the heuristic repair method significantly improves the quality of the solution by decreasing the upper bound when solving large-scale problems.

24 November 2022
Canchen Jiang (Monash University)
Value stacking of EV participation in the power energy system  (view video)

The wide adoption of electric vehicles (EVs) enables technologies of vehicle-to-home (V2H), vehicle-to-grid (V2G), and energy trading among EVs in the distribution network. The coordination of EVs can provide value to the grid and generate benefits for EVs but is subject to local network constraints. This work develops a value-stacking optimisation problem maintaining local network constraints to maximise the value of EVs, considering deterministic and stochastic scenarios. In the deterministic scenario, we assume that we can perfectly forecast all parameters in the energy system, such as solar PV generation and load demand consumption. We optimise EV scheduling, especially discharging, to leverage the multiple value streams, including V2G, V2H, and energy trading among EVs, to minimise the cost of prosumers' daily energy usage. The simulation results demonstrate that our value-stacking model achieves significant cost reductions in Australia's National Electricity Market (NEM), ISO New England (ISO-NE), and New York ISO (NY-ISO) in the US. For the stochastic scenario, in the initial step, we develop a multi-stage stochastic optimization model to improve the decision-making process of value-stacking under load demand uncertainty. Additionally, we utilise the Stochastic Dual Dynamic Programming (SDDP) algorithm to calculate the optimal value-stacking profile for minimising all prosumers' daily operation costs. In the future, we will consider EV arrival and departure time as uncertainties in the energy system.

17 November 2022
Dr. Firouzeh Taghikhah (Univ. Sydney Business School)
Integrated modeling of extended agro-food supply chains: A systems approach (view video)

The current intense food production-consumption is one of the main sources of environmental pollution and contributes to anthropogenic greenhouse gas emissions. Organic farming is a potential way to reduce environmental impacts by excluding synthetic pesticides and fertilizers from the process. Despite ecological benefits, it is unlikely that conversion to organic can be financially viable for farmers, without additional support and incentives from consumers. This study models the interplay between consumer preferences and socio-environmental issues related to agriculture and food production. We operationalize the novel concept of extended agro-food supply chain and simulate adaptive behavior of farmers, food processors, retailers, and customers. Not only the operational factors (e.g., price, quantity, and lead time), but also the behavioral factors (e.g., attitude, perceived control, social norms, habits, and personal goals) of the food suppliers and consumers are considered in order to foster organic farming. We propose an integrated approach combining agent-based, discrete-event, and system dynamics modeling for a case of wine supply chain. In practice, our proposed model may serve as a decision-support tool to guide evidence-based policymaking in the food and agriculture sector.

Firouzeh Taghikhah is a lecturer in the Discipline of Business Analytics at the University of Sydney Business School. Her research interests are at the interface of computational/complex systems modeling, artificial intelligence, and socio-environmental science to guide evidence-based policymaking.

20 October 2022
Prof Sardar Islam (Victoria University)
Operations Research methods for computer science research and applications (view video)

13 October
Dr Yuan Sun (School of Computing and Information Systems, The University of Melbourne)
Problem reduction based on machine learning for combinatorial optimisation (view video)

6 October
Dr Daniel Reich (US Naval Postgraduate School)
Using Machine Learning to Improve Public Reporting on U.S. Government Contracts  (view video)

The U.S. Government procures more than $500 billion annually in goods and services on public contracts, which it classifies using a hierarchical product and service taxonomy. Classification serves several purposes including transparency in the use of taxpayer funding, reporting, tracing and segmenting government expenditures, budgeting and forecasting. Government acquisition personnel have historically performed these classifications manually, resulting in a process that is time-consuming, error-prone and offers limited visibility into government purchases. Using almost 4 million historical data records on governmental purchases, we improve classification by leveraging natural language processing and machine learning techniques to generate a descriptive manual and a predictive classifier. Our machine learning models are embedded in multiple software applications, including a web application we developed, used by government personnel and other contracting professionals.

Daniel Reich joined the Naval Postgraduate School in Monterey, California in 2018 after holding several positions at Ford Motor Company, ranging from research to management. He was nominated for a rotational leadership program, through which he served in managerial roles in manufacturing, marketing, product planning, and information technology. Daniel holds a BS in applied mathematics from Columbia University in New York and a PhD in applied mathematics from the University of Arizona in Tucson.

29 September 2022
Dr. Sanath Kahagalage (Capability Systems Centre, UNSW Canberra)
An application of exploratory modelling and analysis in defence resource planning and asset management under deep uncertainty (view video)

Budget constraints, conflicting stakes, and political situations make decision-making a difficult task even under the most favourable circumstances. Therefore, every reduction in uncertainty is more than welcome.  On the other hand, substituting assumptions for deep uncertainties might simplify choices in the short term, but the consequences may come at a much higher price in the longer term. As decisions that ignore deep uncertainty ignore reality, this questions the reliability and effectiveness of actions developed using such approaches. Moreover, the handling of the uncertainty by decision-makers, including risk managers, is in question. We show the consequences of ignoring uncertainties by using the concepts from Robust Decision-Making (RDM).  The concepts from Behaviour-Based Scenario Discovery (BBSD) and Feasible Scenario Space (FSS) are also utilized to show the importance of giving a complete picture to the decision-maker.

Sanath Kahagalage is a Research Associate at the Capability Systems Centre, UNSW Canberra. He is an honorary fellow in the University of Melbourne. He obtained his PhD from the University of Melbourne and MSc in applied mathematics at Texas Tech university.

15 September 2022
Simon Dunstall (CSIRO Data61)
Managing wildfire risk for agricultural and horticultural regions (view video)

Wildfires (bushfires and grassfires) can destroy crops, livestock, farm infrastructure and the lives and livelihoods of individuals, businesses and communities impacted by a fire event. In this context it is natural to look to methods that can be used to reduce fire occurrence, the chances of fire escalating to major events, and/or means of halting fire progress so as to protect agricultural lands and/or key supply chain assets. From an OR perspective, these are analytics and resource-constrained optimisation questions.

25 August 2022
Dr Reena Kapoor (CSIRO Data61)
Optimisation and Movement Analytics (view video)

Presenting findings from a major literature view just completed by a CSIRO team into questions associated with extracting information and insight from movements of people and objects captured by tracking systems in industrial, agricultural and environmental contexts.

18 August 2022
Dr. Kyle Harrison (UNSW Canberra)
Project Portfolio Selection and Scheduling for Future Force Design (view video)

A common problem faced by organizations is how to select and schedule an optimal portfolio of projects subject to various operational constraints, such as a limited budget. This problem is known as the project portfolio selection and scheduling problem (PPSSP). In the context of defence, the PPSSP arises as part of the future force design (FFD) planning task, where the objective is to maximise the delivery of defence capabilities through project selection and scheduling. However, addressing the PPSSP in the context of defence has its own unique challenges and nuances, such as long planning horizons and limited availability of (public) data. This talk will discuss the development of two models for the PPSSP, namely a project-oriented model and an option-oriented model, with a focus on how they can be used to address the FFD task. This talk will also briefly discuss the use of meta-heuristic optimisers to provide solutions to the proposed models.

Kyle is a Research Associate with the School of Engineering and Information Technology at the University of New South Wales (UNSW) Canberra, Australia. Previously, he was a Postdoctoral Fellow at the University of Ontario Institute of Technology (Ontario Tech University), Canada. He received his PhD in Computer Science from the University of Pretoria, South Africa, in 2018, and the M.Sc. and B.Sc. degrees in Computer Science from Brock University, Canada, in 2014 and 2012, respectively. His research interests include computational intelligence, self-adaptive optimisation, fitness landscape analysis, operations research, and real-world applications of complex networks. He has co-authored numerous publications in top-tier journals and conferences and serves as an Editor for the journal Engineering Applications of Artificial Intelligence.

2021 Series

21 October 2021
John Hearne, ASOR Ren Potts Medal recipient 2020
The reserve design problem under climate change

7 October 2021
Juan Calle Salazar, Deakin University
An optimisation journey. The story of Tactix, an optimisation platform to support strategical-tactical decisions at TDM Transportes

Each optimisation project has a story behind it. This presentation shares the story behind Tactix, a platform built to support the strategical and tactical decisions at TDM Transportes, an innovative logistic service provider in Colombia. At the heart of Tactix, there is a powerful space-time network framework that supports the formulation of an optimisation model. The model captures the most important details of TDM transport operations and delivers important insights to the TDM operations team. Some of the insights the model has provided have been counter-intuitive, and this presentation will share a case in which a customer that seemed to be very profitable in the network, in fact was not.

Juan Calle is Ph.D. candidate at Deakin University. Before pursuing his Ph.D., he was a Senior Operations Research Analyst at TDM Transportes, a Colombian based Transportation Company. Previously, Juan worked at Decisionware, a Colombian based mathematical programming company. Juan has taught Operations Research courses in leading Latin American Universities. He has a bachelor’s degree in Industrial Engineering and a master’s in Systems Engineering from the Universidad Nacional de Colombia. He is the cofounder of UNGIDO, the Operation Research Group at the National University of Colombia, and ASOCIO, the Colombian Operations Research Association.

23 September 2021
Asghar Moeini, 2020 ASOR Rising Star Award Recipient
The Sparse Travelling Salesman Problem

9 September 2021
Simon Dunstall, CSIRO
Optimising power system shutdown criteria to reduce wildfire risk

Electricity distribution businesses in bushfire prone areas have to manage their networks carefully in order to reduce their chances of igniting major fires. Turning off the electrical supply when the fire weather conditions are most dangerous is one of the risk reduction methods at their disposal. Doing so is an action that comes at substantial cost for the community and which also introduces new risks, so it is a tool to be used intelligently and rather sparingly. Data science can be used to quantify the costs and benefits at particular times at specific parts of the network, and optimisation can be used to select the conditions under which the power system is preemptively shut down. The development of these mathematical approaches and their application to statewide electrical distribution networks is the subject of this seminar.

26 August 2021
Dr. Marcella Bernardo, U. Wollongong
Bi-Objective Optimization Model for the Heterogeneous Dynamic Dial-a-Ride Problem with No Rejects

This work proposes a bi-objective mathematical optimization model and a two-stage heuristic for a real-world application of the heterogeneous Dynamic Dial-a-Ride Problem with no rejects, i.e., a patient transportation system. The problem consists of calculating route plans to meet a set of transportation requests by using a given heterogeneous vehicle fleet. These transportation requests can be either static or dynamic, and all of them must be attended to.

12 August 2021
Phil Kilby (CSIRO Data61), reporting on joint work with Dan Popescu and Steven Edwards
Finding solutions to a 3D packing problem arising in logistics

Packing problems have been widely studied in O.R - from cutting carpets to filling knapsacks. Given that, it is surprising how little work has been done on how to fill a truck. We will describe the classic 3D packing problem, and a restriction we have been investigating that arises as a subproblem when solving problems in logistics. While this is still early work, the methods used may be of interest to others.

29 July 2021
Harry Gielewski
Validation to Manage Model Risk

15 July 2021
Dr. Ripon Chakrabortty (UNSW/ADFA)
Merging Data Analytics and Decision Analytics towards Project Management Roadmap: Future Perspectives

1 July 2021
Dr. Ismail Ali (UNSW ADFA)
A novel differential evolution mapping technique for generic combinatorial optimization problems

17 June 2021
Reena Kapoor and Rodolfo García-Flores (CSIRO Data61)
Optimal Schedules for Corn Planting and Storage

Corn (or maize) is, with rice and wheat, one of the most consumed cereals in the world, together accounting for 94% of all cereal consumption. It is estimated that, in 2012, the total world production of corn was 875.23 million tonnes. The development of seeds with desirable traits typically requires many years of in-field testing before new products can be delivered to market. Recently, innovative genomic technologies have shortened the time required to develop new corn hybrids, that is, new products that can deliver higher-yielding, better-adapted seed options for growers at a faster pace. However, higher yields and increased rates of produced parental lines introduce many new challenges. In this presentation, we address one such challenge, namely, the problem of managing the demands on storage facilities to cope with increasing output (i.e., the number of harvested ears). The problem was proposed by Syngenta Seeds to improve their year-round breeding process by optimizing planting schedules to achieve a consistent output, which translates into a weekly harvest quantity. Erratic weekly harvest quantities create logistical and productivity issues. The research question we address is: How can we optimally schedule the planting of our seeds to ensure that when ears are harvested, facilities are not over capacity, and that there is a consistent number of ears each week? The solution we present is the winner of the 2021 Syngenta Crop Challenge in Analytics.

2020 Series

10 December 2020
A Mathematical Modelling Approach for Managing Sudden Risk in Supply Chain (view video)
Sanjoy Paul, University of Technology Sydney, and 2018 ASOR Rising Star Award Recipient

3 December 2020
A matheuristic solution approach for the p-hub center and routing problem over incomplete hub networks (view video)
Zühal Kartal, Eskisehir Technical University, Turkey

19 November 2020
Gurobi v9.1 (releasing this week!) capabilities, new features and performance (view video)
Sebastian Thomas, Account Director – Oceania and Southeast Asia, Gurobi Optimization, LLC
and Kostja Siefen, Technical Account Manager, Gurobi

12 November 2020
Resources and methods for fire risk analysis (view video)
Simon Dunstall (CSIRO Data61)

5 November 2020
Lessons learnt from COVID-19 surge modelling for the Australian Royal Flying Doctor Service (view video)
Hannah Johns (Florey Institute / U. Melbourne)

10 September 2020
Online Incentive-Compatible Mechanisms for Traffic Intersection Auctions
David Rey (UNSW, Research Centre for Integrated Transport Innovation)

3 September 2020
Simulation-based optimization in fleet management
Hasan Turan (UNSW/ADFA)

27 August 2020
The use of the Sports Synthesis model to determine appropriate draft penalties in an AFL-like sports league
-- Recorded presentation is hosted on vimeo at https://vimeo.com/517668131
Geoff Tuck (CSIRO Oceans and Atmosphere, Hobart)
- Geoff will describe how a simulation model can be used to quantify the impact on success of alternative draft penalties for a club that has breached league regulations – and how this relates to fishing.

13 August 2020
A decomposition framework for capacity expansion planning with unit commitment
-- Recorded presentation is hosted on vimeo at https://vimeo.com/517667756
Semini Wijekoon (Monash University)
- Development of an approach for solving electricity network design problems that is derived from scenario decomposition (SD) techniques.

6 August 2020
Multiperiod storage system modelling in the context of nonlinear power network optimisation
-- Recorded presentation is hosted on vimeo at https://vimeo.com/517667505
Frederik Geth (Research Scientist, CSIRO Energy)
- Combining multiperiod storage models and power flow physics results in large nonlinear optimisation problems. The talk discusses the application of recent reformulation techniques and implementation aspects in Julia/JuMP/PowerModels.

30 July 2020
Hyper-heuristic for Combinatorial Optimisation Problems
-- Recorded presentation is hosted on vimeo at https://vimeo.com/517667192
Ayad Turky (Victoria University)
- Hyper-heuristic (HH) is a high-level search methodology that searches for problem solving methods, rather than problem solutions, and can be succesfully applied in resource allocation, scheduling, routing, production planning and economic systems.

23 July 2020
Performance analysis and feasibility of hybrid ground source heat pump systems in fourteen cities
-- Recorded presentation is hosted on vimeo at https://vimeo.com/517666880
Hansani Weeratunge (U. Melbourne)
- Hansani has recently completed a PhD applying simulation and optimisation to the design and operation of energy-efficient building heating and cooling systems based on ground-source heat pumps.

16 July 2020
An Application of Business Rule Optimisation
-- Recorded presentation is hosted on vimeo at https://vimeo.com/442309070
Alan Dormer (Opturion P/L and Monash University)
- Alan's recently awarded PhD is on the optimal choice of business rules for recurring decisions in domains such as customer service, finance and health.

9 July 2020
Big Data Analytics and Machine Learning for Smart Cities
-- Recorded presentation is hosted on vimeo at https://vimeo.com/437779815
Peter Ryan (Honorary Research Fellow, Defence Science & Technology Group) and Richard Watson (Research Scientist)
- Peter, Richard and colleagues have been looking at the application of analytics to cities, with an emphasis on open data sets provided by City of Melbourne.

2 July 2020
Simulating the spread of COVID-19
-- Recorded presentation is hosted on vimeo at https://vimeo.com/437015924
Phil Kilby (Principal Research Scientist, CSIRO Data61)
- Phil is part of a team based in CSIRO and Department of Health which is looking at the dynamics of COVID-19 outbreaks and our response to keeping these outbreaks under control. 

25 June 2020
Predicting solutions of large-scale optimization problems via machine learning: A case study in blood supply chain management (view video)
Zahra Hosseinifard (Lecturer of Operations Management, Faculty of Business and Economics at The University of Melbourne)
- Zahra will be talking us through her recently-published co-authored paper in Computers and Operations Research.



 

A selection of older ASOR Seminars

2 June 2017 - Nicholas Davey

The ASOR Melbourne AGM on 2 June 2017 was preceded by a seminar Optimal road design through ecologically sensitive areas considering animal migration dynamics delivered by Nicholas Davey (University of Melbourne).

30 September 2015 - Andreas Ernst

A retrospective of over two decades of Operations Research at CSIRO

Abstract: The Operations Research group at CSIRO was formed in the early 1990s and since then has been a significant part of the OR scene in Australia. With the departure of a number of key members of the group and the imminent merger between NICTA and the CSIRO, the OR group will cease to exist in its current form. This talk will look back over the past 20 years at some of the highlights of what has been achieved by the group. This includes discussion of how a small project for Australia Post inspired a stream of research into hub location problems; the development of rostering optimisation methods and the challenges of commercialising it; and the range of interesting OR problems in bulk material (mining) supply chains. The talk will also be used to comment on the developments in OR in Australia more widely and the likely trends into the future.

23 September 2015 - Asef Nazari

Expansions on Land-use Trade-off Optimisation (LUTO)

Abstract: CSIRO has previously developed a model of land-use trade-offs that considers the possible evolution of agricultural land areas in Australia over the next 40 years. This can be modelled as a large scale multi-stage linear programming problem. However, acquiring the expected outcome requires solving the large scale LP problem which takes more than one hour to solve for a single year. In this regard, we developed a combination of aggregation-disaggregation technique with the concept of column generation to solve the large scale LP problem originating from land use management in the Australian agricultural sector in a shorter amount of CPU time. In addition, increasing  demand for greener energy alternatives are putting more pressure on the use of agricultural land for not just food productions but also biofuels, carbon sequestration, biodiversity and other non-traditional uses. A key question is how this is going to impact not only the land use but also the agricultural supply chains that process the outputs of the land use. In this talk we also initiate the question of locations of processing centres and land use in an integrated optimisation model. Here we consider in addition the construction of some processing centres for bio-fuel, bio-energy, livestock facilities and so forth, which introduces a new combinatorial aspect to the model. The decisions of land use and the location of processing centres are interlinked as transport costs based on distances are often instrumental in determining the economic viability of some of the land uses and conversely economies of scale are necessary to justify investment in processing plants. We introduce a model containing both problems of a land allocation and a facility location simultaneously which results in a large scale mixed integer linear programming (MILP) problem and therefore is computationally difficult to solve, and we will cover some of the computational difficulties.
 
Biographical Info: Asef was awarded his PhD in 2009 on the topic of developing derivative free algorithms for non-smooth optimisation problems from the University of Ballarat. Immediately after his PhD, he was appointed as a research associate at the UniSA to conduct research on the optimal expansion of a power system. Since 2013 he has been employed by CSIRO to be involved in several industrial projects

2 September 2015 - Kristian Rotaru

3.30pm, Room 7.84, Building H, Monash Caulfield

Risk information processing and decision-making with strategic performance measurement systems: an eye-tracking study

To address the limitations of the traditional strategic performance measurement systems (SPMSs) in visualizing risk and preventing excessive managerial risk-taking, a number of research studies proposed to extend the functionality of SPMSs by incorporating risk information into traditional SPMSs, such as balanced scorecards. Thus, despite the growing calls of practitioners and researchers on combining performance and risk measures as part of an extended framework, there is a lack of uniform vision about what constitutes such a framework. The aim of this study is to investigate how the representation of risk-related information (characteristics of risk events and key risk indicators) in SPMSs influences the identification and processing of this information in managerial risky decision making. This study benefits from the use of eye-tracking methodology in a laboratory experiment, which allowed to acquire better understanding of the cognitive processes and the subsequent behavioural response associated with managerial risky decision-making when using SPMSs as a tool for decision support.

Dr Kristian Rotaru (PhD in Economics, PhD in Information Systems/Risk Management) works in the domains of risk modelling and decision making. He is a Member of the Editorial Review Board of the Journal of Operations Management. At Monash Business School, Kristian leads the Risk Analysis, Judgement and Decision-Making cross-disciplinary research team that focuses on integration of normative research informed by analytical and simulation modelling methods and descriptive research, informed by laboratory and field experiments. In his research he adopts a variety of research methods, including market data analysis, conceptual, analytical and simulation modelling and laboratory experiments (involving the use of eye-tracking and electroencephalography technology). Kristian lectures Business Analytics, Accounting Information Systems and Financial Modelling units.

19 August 2015 - 2pm - Carleton Coffrin (NICTA)

Carleton gave us an entertaining and insightful presentation about solving the Optimal Power Flow (OPF) problem for AC electrical power networks. In OPF decisions are made about the amount of electricity generation undertaken at generation nodes in a network, so that demand is fulfilled at a series of demand nodes - subject to the non-linear and non-convex constraints relating to AC power flow in transmission networks. There are successive relaxations to the full problem: via semi-definite programming, conical programming, a transport model, and finally a "copper sheet" model without transmission line flow limits. On standard sets of test instances the strongest two relaxations displayed remarkable performance, i.e., finding optimal every time... but further investigation showed that these standard instances were in fact too easy to solve to global optimality, because the data was such that certain sets of constraints would never be active. This led to the development of better benchmark datasets which have proven far more interesting to solve and which are true tests of algorithms for Optimal Power Flow.