天基预警调度的启发式优化方法

天基预警过程可以看作一种多维离散时间序列监控与预测问题,其调度的决策要素、优化目标和约束条件较多,故往往采用智能优化算法求解该非线性优化问题.而它们在指定时间内却是概率性收敛到Pareto解集.对此,提出基于贝叶斯方法提供多类别决策树挖掘调度中的启发信息,以及引入局部搜索算子等方法提高智能优化算法的快速性和鲁棒性.预警仿真实验表明融入上述方法的免疫克隆选择算法收敛性能提高了10.1%,遗传算法提高了9.8%.

Heuristic optimization of resource allocation in space based early-warning system

Space based early-warning is the monitoring and predicting process with the property of the discrete time sequence. Many decision-making factors, optimization objective functions and constraint conditions need be considered when scheduling the sensor tasks, as a result, the intelligent optimization algorithm is typically adopted to solve this nonlinear optimization problem. However these optimization methods are convergence to the Pareto solution in probability within the specified time. Three respects of work are done in this paper, firstly, the Decision Tree is applied to mine the heuristic knowledge; secondly, the Smoothed Naive Bayes is presented in order to provide the multi-class Decision Tree; finally, the heuristic knowledge is fused into the intelligent optimization algorithm, where local search operator is added. The experiment shows that Immune Colonal Selection increases 10.1% in terms of convergence probability, and Gene Algorithm increases 9.8%.