Optimizing tree planting areas through integer programming and improved genetic algorithm

The simulation of planting lining design relative to possible solutions of dividing an area to blocks (block division) and selecting planting lining direction towards optimizing tree planting areas is a complex problem. To analyze the huge number of possible solutions with uncertain results manually will lead to unanswered decision of the optimal result determination. Therefore, a hybrid algorithm through an incorporation of Integer Programming and Improved Genetic Algorithm was proposed for planting lining design. The algorithmic solution with some strategies mainly focuses on efficiency. Planting lining design selection in oil palm planting areas involves the comparison of the three techniques namely technique-1 which is to handle 600 planting direction from baseline, technique-2 which is to choose the best planting lining direction in an area and technique-3 which is to obtain optimal block division. Then the best planting lining directions are assigned into the blocks. The decision based on the highest number of trees is promoted among the three techniques. The process of block division and determining the optimal number of trees require a series of analysis. The possible solutions rely on the number of blocks or the number of shapes that represent the blocks, where an increase in both numbers influences the rising time for analysis. Previous strategies by cell representation promote excessive time to generate solution for large areas, whereas two methods of best-fit called Bottom Left to Right first and Bottom Left to Top first promote result inconsistency and require more time in analyzing solution. Therefore, the two strategies consisting line representation and combined best-fit methods were introduced to solve two issues respectively; the issues are to count tree number according to planting lining directions and to decide block division. Meanwhile, the improvement in Genetic Algorithm is focused on the strategies of specific random value, deterministic crossover and deterministic mutation. In addition, the strategy of control mechanism was applied in hybrid algorithm. With the aim of evaluating the algorithm efficiency, comparisons between the proposed strategies and the previous strategies were conducted. The result of line representation promoting less iteration numbers indicates that time usage is more efficient. More number of optimal solutions in combined best-fit methods and the rejection of infeasible chromosomes in Genetic Algorithm are significant factors to be better efficiency. Moreover, the implementation of control mechanism in order to skip the expected same solution occurs, expedited the processing time. These proposed strategies were applied in an application named the Lining Layout Planning by Intelligent Computerized System. This application generates possible solutions for planting lining design. Analysis results from coordinate dataset shows that the selection of planting lining techniques for areas less than 10 hectare is difficult to predict but for the larger areas, results showed that the technique-3 is more consistent to produce the best tree density. While, the results of FELCRA dataset show through technique-2 or technique-3 produces higher number than common practice which allows 143 trees/ha. In conclusion, the hybrid algorithm based solution strategies improved efficiency with convincing results, therefore, this will assist planners for better decision making to optimize area to achieve more trees to be planted. The model solution and empirical result are important for the selection of planting lining design. The experimental results on the datasets within the actual coordinates of areas are not only confirmed as theoretical results but also prove that it can be applied in practice. In addition, the revealed results contribute to the new perspective of designing planting lining for area optimization by computerized system.

Preview PDF FSKTM 2012 6R.pdf Download (1MB) | Preview

Actions (login required)

View Item