Apart from hospital and healthcare personnel scheduling, there are several other organisations which require personnel attendance 24 hours per day. Glover and McMillan [105], for example, define a general employee scheduling problem and tackle the design of shifts and the assignment of these shifts to
Burns and Carter[45], (1985), present an operations research approach for employee scheduling. The main aim of the work is to design work cycles which minimise workforce requirements and fulfil a set of constraints. Some of the presented heuristics for solving the problem are proven to be optimal. The constraints treated by Burns and Carter are a subset of the constraints presented in Section 2.4, only the simplest constraints are considered. Emmons and Fuh
[90], (1997), developed a method for optimising the workforce, and for scheduling full-time and half-time personnel while considering some soft constraints. Millar and Kiragu [145], (1998), combine all the possible shift patterns of 4 days length. They construct a network in which each node represents a feasible pattern and solve it with an algorithm based on CPLEX software.
Although doctors and surgeons are hospital personnel, they are normally not scheduled in the way nurses are. In some cases, a separate algorithm is developed for this category of healthcare personnel (e.g. Graff and Radford [110]). Locations (operating theatres, patients’ addresses, for example) or avail- able equipment
such as cars (De Causmaecker et al. [74]), specialised medical equipment (Schreuder [185]) can play an important role in the healtcare sector. Similar problems are those of magistrates in justice courts (Schreuder [184]).
Many of the employee scheduling problems are less complex than hospital scheduling, e.g. the bank sector, customs personnel, call centers, postal centers, etc. They are often solved with cyclical schedules, which are also very common in production environments. Media personnel, for example in broadcasting stations and publishing environments, cannot work according to perfect cyclical patterns either.
All kinds of commercial activities require employee scheduling: sales assistants, cash registers in stores, telephone sales, etc. Fast food restaurants form another group, in which duties are composed of very short tasks. Personnel requirements per task depend strongly on the time of the day. Good schedules are those in which people can work continuously, even though they may be undertaking different tasks all the time.
The personnel demands in many other personnel scheduling, are fluctuating less from day to day and the number of different skill categories for personnel is generally lower. However, some other personnel timetabling domains include constraints which are normally not considered in hospitals. Examples are law enforcement (Tien and Kamiyama [204]), prison staff, police (Taylor and Huxley [202]), security personnel, fire departments, service personnel (Collins and Sisley [59]), etc in which locations can play a role. Such schedules must address emergency situations in addition to routine personnel demands. In healthcare, emergencies are often covered by assigning the typical ‘keep guard’ duties. The personnel members which are on guard duty are called when a sudden personnel shortage occurs. Military manpower scheduling problems form a special group within this category.
programme and on rooms with special facilities make the problems very hard to solve.
Other related problems which involve locations are audit personnel scheduling, generating preachers’ timetables (Corne and Ogden [62]), etc. Courier services, telephone engineers, transportation (B¨urckert et al. [32]), personnel at airline stations [31], sanitation (Tien and Kamiyama [204]), etc incorporate constraints on vehicles and routes. The problem definition differs strongly from the subject of this thesis.
There is a group of problems called ‘crew scheduling’ (Beasley and Cao [13], Bianco et al. [22], Caprara et al. [47], Morgado and Martins [151]): scheduling buses (Wren and Rousseau [225]), with e.g. partial driver shifts, trains, boats, airports and aircrafts (Dowling et al. [82], Gopalan and Talluri [109]), etc. A team of people, with the required skills, has to take a vehicle from one place to another. In many situations, locations cause extra difficulties when the crew stays at the destination point until another vehicle is scheduled back.
The personnel rostering techniques which are developed for this research are to a certain extent applicable to other personnel scheduling domains but in many cases, a lot of extra information is required.
Some scheduling approaches for non-healthcare personnel have been intro- duced in Section 3.2 and Section 3.3 already [30, 55, 85, 138, 140, 154]. A few other problems, which differ even more from the nurse rostering problem, are presented in this section.
Bailey[9], (1985), models days on and off patterns for service organisations with hourly fluctuating demands. He distinguishes ‘shift scheduling’ and ‘days off’ scheduling; the first determining the number of 8-hour shifts needed to satisfy fluctuating demands during the day and the second one to determine the weekly work patterns for personnel members. Bailey presents a decom- posable linear programming approach. Constraints on the daily level, such as consecutive days off, staff size, overtime, etc belong to the shift scheduling part of the problem while hourly demand variations, changes in the cost of understaffing and overstaffing are reflected in the days off problem. The goal is to minimise the number of work patterns with the highest difference between start times over the week.
Alfares [6], (2000), developed a method to solve a real-world -non healthcare- employee timetabling problem with days on and off, which is especially suitable for personnel working in remote areas. He minimises the number of assigned employees while respecting a very strict cyclical schedule. The presented method is optimal, it makes use of the dual LP formulation but avoids the inefficient use of integer and linear programming. The application is used by an oil company to schedule workers in remote areas.
to tasks and locations in the form of shifts under flexible workload conditions. The method aims at a long-term fair distribution of undesired shifts. Different contracts involve different values for the constraints. A planning period of one week is common. The qualifications of employees enable them to fulfil certain types of tasks.
In the objective function, the following soft constraints are implemented as a weighted sum:
- preferential ability and availability: personnel members express their pref- erences for certain shifts and for work requiring particular skills
- flexible workload: violations on the number of assignments per week (both positive and negative) have to be evenly distributed among personnel members and over time
- fairness for special shifts (can be achieved with counters, Constraint 23) - shift and location stability: people prefer to be assigned to the same shift
and the same place in a week time (in ANROM, locations are not a subject to consider; the shift stability could be attained by setting the minimum value for consecutiveness of certain shifts, Constraint 5).
Special shifts can be defined by the user, typical examples are weekend shifts and night shifts. For test examples with 50 and 100 employees, the method generates good solutions and requires very few calculation time.
Cowling et al. [64], (2000), test several hyperheuristics on a sales summit scheduling problem. Hyperheuristics do not require much problem specific knowledge and therefore represent a very promising research direction for building more general scheduling systems. A choice function determines which low-level (local search) heuristic to choose from a set of problem-independent algorithms, under the given circumstances.
The problem consists of organising meetings between ‘suppliers’ and ‘delegates’, subject to constraints. Although this problem does not resemble the nurse scheduling problem at all, the hyperheuristic approach certainly has potential to be applied for different types of personnel scheduling problems.