Abstract—Nurse scheduling is a difficult problem that every hospitals faces every month or a week. It impacts the health care system in the hospitals. The Nurse Rostering Problem [NRP] said to be as a subclass of the personnel scheduling problems, and most of its instances are not solvable. Solving the Nurse Rostering Problem has been research field from many years, though, Nurse Rostering is done manually. This paper provides the information about various methodologies for solving NRP and tells NRP can be solved by GA and PGA with the help of review.
Index Terms- GPGPU (General Purpose Graphics Processing Units), Constraint Programming, Soft Constraints, Hard Constraints, Genetic Algorithm, Heuristics, Nurse Rostering Problem, Parallel Genetic…show more content… Flowchart Of Genetic Algorithm [12]
Using mutation and crossover operators new offsprings (solutions) are created from the current population. Populations are interchanged following the selected population model until the specified ending criterion is met.
The population of solutions forces the genetic algorithm to be more computationally complex, when compared to other heuristic methods. On the other hand a population of solutions also gives the genetic algorithm better robustness, which turns out to be a advantage especially when dealing with problems where the fitness function contains many local optima[13]-[15].
IV. PROPOSED PARALLEL APPROACH
In PGA, all the steps of Genetic Algorithm are performed but the fitness function evaluation is performed on GPU and results are sent back to GPGPU. Figure below shows the flowchart of parallel genetic…show more content… Flowchart Of Parallel Genetic Algorithm In [16] paper, author has proposed an implementation of steady-state GA on a GPU using CUDA, and also adopted the data parallelization on GPU architecture. Experimental results have shown that, the proposed implementation method achieved approximately 6 times faster results than implementation on CPU. But there is need to improve search performance in the steady-state model. There is also need to adopt different tasks that have non uniform computational granularity so that improved steady-state models will be more suited to the GPU environment [16].
An image matching algorithm based on genetic algorithm on GPU is implemented by Yongzhen Ke [17]. Designed fitness function, chromosomal selection function, crossover, and mutation function are suitable for running on GPU. The experimental results have proved that the proposed algorithm can match optimal result quickly and achieves satisfying speedup under the conditions of the existing