Variable Neighbourhood Search: A case study for a highly-constrained workforce scheduling problem

Abstract

Abstract-This paper describes a Variable Neighbourhood Search (VNS) combined with Metropolis-Hastings acceptance to tackle a highly constrained workforce scheduling problem typical of field service operations (FSO) companies. A refined greedy algorithm is firstly designed to create an initial solution which meets all hard constraints and satisfies some of the soft constraints. The VNS is then used to swap out less promising combinations, continually moving towards a optimal solution until meeting finishing requirements, which are either a satisfactory mean fitness set as a parameter, or a time allowance of one hour. The results of this approach are promising when compared to the stand-alone greedy algorithm.

A novel encoding for separable large-scale multi-objective problems and its application to the optimisation of housing stock improvements

“…For this to be the case an application will be like the housing stock problem where there are two scales: sub-problems where there is a trade-off, but these subproblems interact at the global scale. Another example is employee scheduling [13], where regions might be scheduled to both meet the target workload and minimise the required workforce; but increasing the workforce in one region will require a reduction in the workforce for another. Identification and classification of such problems is an interesting direction to pursue.…”

“…An interesting property of housing stock optimisation is that both the cost and energy objectives have the same separability: the problem for each house (or group of houses if they share, for example, a district heating system) is independent of the others. This property is shared by employee scheduling for large companies [13], specifically where employees are scheduled at a regional level, and regions are independent of each other. For single-objective additively-separable problems, the search complexity can be reduced by considering sub-problems in isolation [14]; however, for multi-objective problems, we show that there is an additional complication: when considering the objectives together the problem is, in effect, non-separable where the objectives share the same additive separability.…”

Shift Scheduling and Employee Rostering: An Evolutionary Ruin & Stochastic Recreate Solution

“…Both the shift scheduling and employee rostering problems are typical of organizations which can range from a small number of employees such as nurse rostering [17] to larger scale such as field service engineers [13]. This problem deals with scheduling and rostering a large number of engineers (around 25,000), ranging up to around 200 employees.…”