Dynamic rostering keeps industry on track
Researchers
at the University of South Australia have developed an intelligent system
known as dynamic rostering that calculates the best possible work rosters
automatically and dramatically improves efficiency over manual scheduling.
Unlike cyclic rostering, which works on static requirements, dynamic rostering is based on actual schedules, using available staff with the required skill levels, and with fatigue and leave taken into account.
Dynamic rostering guarantees optimal rosters that are fair, transparent and equitable, with no manual changes required, according to Research Fellow Dr Guy Eitzen from UniSA’s School of Mathematics and Statistics, who developed the mathematical algorithms for dynamic rostering.
“The intelligent rostering system is a mathematical optimisation algorithm built into a computer program, which takes all of the information entered and produces a roster that people can work with,” Dr Eitzen said.
With researchers from UniSA’s Centre for Sleep Research, Dr Eitzen has been investigating the best mathematical algorithms to solve multi-skilled rostering problems for a variety of industries including power generation, mining, correctional services, aviation, food production, railway, trucking and traffic control.
In a major research project aimed at keeping the rail industry on track, Dr Eitzen used dynamic rostering to automatically generate the best rosters for train crews, enabling roster clerks to reduce the amount of time performing changes to the cyclic roster. In some cases this could equate to 30 to 40 per cent of their time.
“A peculiarity with trains is that they leave from one destination and arrive in another, compounding crew rostering difficulties. Dynamic rostering for industries based in one location is much easier and works very efficiently in the industries that we have studied,” Dr Eitzen said.
“Most railways around Australia use a cyclic based roster, which means that everyone works the same rosters over time, routinely working on each line of the roster before returning to the first.
“Cyclic rostering works well if there are no changes, but if people work more or less than their expected hours and are not working to the same schedules as other employees, there’s an equity issue and the potential for fatigue levels to be inadequately addressed.
“In many cases cyclic rostering ignores skill levels and leave days, it takes into account the expected trains as shown on a master train plan (usually prepared months in advance), but not the actual trains running.
“To ensure that enough people with the right skills are available to meet actual requirements, labour intensive manual changes are made to the roster 48 hours in advance, but with no optimisation or equity taken into account. While the actual trains running are known in most cases more than two days ahead of schedule, the cyclic roster is not updated earlier as changes in the week before departure may mean having to redo the roster,” Dr Eitzen said.
In contrast, dynamic rostering is done automatically with no manual changes required, making it faster and much more efficient than cyclic rostering. It can be done at any length of time in advance as long as train schedules are known with a fair degree of certainty. The system can also handle any variations to the master train plan, such as departure time changes, additional train services and cancelled train services.
Predicted days off can be allocated in advance based on forecast rather than actual trains, giving extra levels of predictability over the shorter planning horizon. Because shiftwork is involved, employees can often have work assigned around their requested days off without having to take annual leave, which is a distinct advantage for them.
Employees can also have their preferred working times recorded using a preference score for every hour of the week. By overlaying the rostered hours assigned to workers, the system can rate the roster in the eyes of employees - a low score is a good roster and a high score is bad - and can be used to ensure that each employee works equal amounts of preferred hours. Preferred working times cannot be handled in cyclic rostering.
Using dynamic rostering, work rosters can be reconfigured over and over again until roster clerks are satisfied with the schedule.
The dynamic rostering research is being conducted with a consortium of railways around Australia including Queensland Rail; CityRail, Pacific National and Silverton Rail in NSW; Australian Railroad Group in WA; and Pacific National and Great Southern Rail in South Australia.
”Silverton Rail is implementing dynamic rostering, with planning and processing of data underway. It suits their business really well and they want to develop it further,” Dr Eitzen said.
“Ultimately it will save a lot of time because at the moment cyclic rostering in this industry involves a huge amount of manual work.”