120 / 2023-07-10 21:24:32

无人机集群决策边界自适应采样方法

无人机集群；自适应采样；代理模型；仿真试验,UAV Swarm, Adaptive Sampling, Surrogate Model, Simulated Test

体系试验鉴定与评估 > 1.试验设计与评估方法

针对无人机集群测试场景样本需求巨大，测试时间耗费大的问题，本文提出了一种自适应采样方法来对使无人机集群性能表现发生关键变化的场景进行采样。该算法将无人机集群视为黑盒自主系统，以无人机集群不同测试环境下完成任务的时间作为响应，将测试环境参数大小视为影响因子，拟合高斯过程回归模型；设计可以调节样本接受条件的模型评估器，根据不同采样目的控制采样策略。通过算法示例分析，发现本文所提出的方法可以更快拟合模型以及较好地对决策边界附近区域进行采样，有效减少无人机集群测试场景样本的需求，减少测试试验时间，提高了测试效率。

Aiming at the problem of huge demand for UAV swarm test scene samples and large test time consumption, this paper proposes an adaptive sampling method to sample scenes that make key changes in the performance of UAV swarm. The algorithm regards the UAV swarm as a black box autonomous system. The time of completing the task in different test environments of the UAV swarm is taken as the response, and the size of the test environment parameters is taken as the influence factor to fit the Gaussian process regression model. A model evaluator that can adjust the sample acceptance conditions is designed to control the sampling strategy according to different sampling purposes. Through the analysis of algorithm examples, it is found that the method proposed in this paper can fit the model faster and sample the area near the decision boundary better, which can effectively reduce the demand of UAV swarm test scene samples, reduce the test time and improve the test efficiency.