Current methods that use Unmanned Aerial Vehicle (UAV) swarms to inspect roads still have many limitations in practical applications, such as the lack of or difficulty in the route planning, the unbalanced utilization rate of the UAV swarm and the difficulty of the site selection for the distributed droneports. To solve the limitations, firstly, we construct the inspection map and remove the redundant information irrelevant to the road inspection. Secondly, we formulate both the route planning problem and the droneport site selection problem in a unified multi-objective optimization model. Thirdly, we redesign the encoding strategy, the updating rules and the decoding strategy of the particle swarm optimization method to effectively solve both the route planning problem and the droneport site selection problem. Finally, we introduce the comprehensive evaluation indicators to verify the effectiveness of the route planning and the droneport site selection. The experimental results show that (1) with the proposed method, the overlapped part of the optimized inspection routes is less than 7% of the total mileage, and the balanced utilization rate of the UAVs is above 75%; (2) the reuse rate of the distributed droneports is significantly improved after optimization; and (3) the proposed method outperforms the ant colony optimization (ACO) method in all evaluation indicators. To this end, the proposed method can effectively plan the inspection routes, balance the utilization of the UAVs and select the sites for the distributed droneports, which has great significance for a fully autonomous UAV swarm inspection system for road inspection.
15 January 2013 in International Journal of Digital Content Technology and its Applications
In this paper, we discretized the original data by self organized mapping (SOM) neural network based on the characteristic attributes of the 3rd party logistic providers. Decision table was achieved. The redundant attributes were removed by rough sets (RS) in order to get the simplified decision table. Then we trained the support vector machine (SVM) with the simplified decision table to construct the model that could embody the the mapping relationship between the characteristic attributes and the evaluation grades. The test result shows that the specific characteristic attributes of the 3rd party logistic providers from the same area can be distilled after processing the data with RS. These characteristic attributes' influence on the evaluation grades is determined by the core mined from the original data. Besides, the establishment of SVM model provides a new method that can reduce the calculation work of the grade evaluation on a new logistic service provider. Furthermore, this method can also store and reuse the experts' experiences of evaluation.