Operations Research in CMIS

Operations Research (OR) is about making informed decisions about the planning and operations of large and complex systems when resources are scarce.

The OR researchers in the CSIRO Division of Mathematics, Informatics and Statistics contribute to the Adaptive Supply Networks stream as well as several of the National Research Flagships and the Sensors and Sensor Networks capability platform within CSIRO.

In the past, many of us were part of the Business Decision Tools team, which provided research services in domains such as rostering. Prior to that, we worked as the Operations Research Group from 1992 to 2003, and during this time our engagements with industry clients proved our ability to apply our skills in a diverse range of application areas.

Some Operations Research methods

Simulation

Simulation is a tool that is frequently used by CSIRO Operations Research (OR) specialists when tackling commercial problems. There are many simulation case studies to be found amongst our descriptions of client work.

Combinatorial Optimisation

Better scheduling, planning and rostering is made possible through the application of methods for combinatorial optimisation. Reliable construction of a good timetable, roster or schedule depends critically upon the selection of a good combination of activity timings and resource allocations from an almost limitless set of possible solutions. The massive size of this set makes it impractical to directly evaluate even a small proportion of the candidates. However, our application of various combinatorial optimisation methods, such as simulated annealing and tabu search, enables us to home in on optimal or near-optimal solutions to real-world problems in realistic computational times.

Linear, Integer and Mixed-Integer Programming

These mathematical programming methods can allow large-scale continuous or combinatorial optimisation problems to be solved using custom-built mathematical formulations and standard solvers. Operations Research specialists make use of solvers such as CPLEX in both commercial work and scientific research. For example, for commercial work in train crew rostering we used a clever MIP formulation and a CPLEX runtime library bundled together in a software application.

Queuing Theory

Projects such as our work with RTA require us to look at the formation and behaviour of queues. The understanding, planning and engineering of queues is often of critical importance when working in domains such as customer service centres and demand-driven manufacturing. Analyses of a real-world queuing systems frequently reveal startling information about the crucial dependencies between arrival rates, servicing rates, queuing system design, and target service standards.

Page last updated February 03, 2010 12:46 PM