Research On Route Filtering Method Based On Pareto-optimal Analysis

As one of the cores in task planning of the unmanned aerial vehicle,the main function of route filtering is to further prune the existing route set,so as to leave the routes that users may be interested in,and reduce the range of choice for decision makers in the final decision.For conventional route filtering algorithm,decision makers tend to artificially assign weight or relative importance for each filtering criterion on their own,transforming the route filtering problem into a single-objective optimization problem to solve.although to a certain extent,it makes the complex route filtering problem simplified,but also causes that route quality is heavily dependent on the proportion between each filtering criterion.In order to effectively solve the defects of traditional route filtering algorithms,in this thesis we propose Adaptive Geometric Angle-based Algorithm for Ranking Pareto-Optimal Routes(Ranked ADA),which not only reduces the knowledge requirements of decision-makers in this field,but also greatly reduces the workload of decision-makers in choosing route.First,we introduce about the multi-objective optimization theory,highlighting the two steps of solving multi-objective optimization problem by implicit method: get the approximate pareto-optimal solution set and filter to get the knee points which have outstanding comprehensive performance.And taking the judging of knee point as the breakthrough point,we point out evaluation standards of route filtering algorithm based on multi-objective optimization analysis: find all knee routes in the Pareto-optimal route set at the lowest cost,effectively solving the traditional route filtering algorithm's dependence on the proportion between route filtering criterion.Then,in this thesis we formally propose the Ranked ADA algorithm based on Adaptive Geometric Angle-based Algorithm for Pruning Pareto-Optimal Solutions.By calculating penalty scores,the set of pareto optimal routes has been ranked in an ascending order,so that the knee routes with outstanding comprehensive performance are placed at the top.In addition,according to the characters of knee point,we put forward the concept of Angle Penalty Score based on spatial geometry theory and Penalty Intensity Ratio based on the Law of the opposite ratio of the square of light.And Then we carried out experiment on standard multi-objective optimization testing function,proving the effectiveness of Ranked ADA algorithm in the search of the knee solution in multi-objective optimization problem,which can find all of the knee points in pareto optimal solution set at the smaller expense.Finally,we carry out experiment of comparison between ADA algorithm and Ranked ADA algorithm in different simulative battlefield environment,demonstrating the four advantages of Ranked ADA algorithm over ADA algorithm:(1)No need to input any parameters.(2)Decision maker can specify the number of routes left after filtering according to the actual application scenario.(3)Improving the clustering phenomenon in ADA algorithm and increasing the diversity of left routes.(4)All knee routes in the Pareto-optimal route set can be found at a lower cost.