盲人探路负梯度方向法

负梯度方向法作为一个常用的优化方法在机械工程领域发挥着重要作用,但是,因其锯齿现象而具有计算量大、计算效率低的缺点。一维盲人探路寻优思想总结为:根据探测点与极值点相对位置的三种情况采取三种处理方案。基于此,将负梯度方向法进行了改进,提出了新的寻优方法——折线负梯度方向法。算法分为四部分:初始步长检验阶段;步长加倍探测阶段;暂不减半步长阶段;步长减半探测阶段。第三部分考虑了探测点远未及极值点的情况。提供了寻优思想流程图和完整的C语言子程序。通过与负梯度方向法的比较,证明了折线负梯度方向法具有计算量小、寻优效率大的特点。考虑远跨过极值点的情况,提出了走一步退半步探的算法。通过对不进行退半步探运算和退半步探时不减半步长两种情况的比较,证明了折线负梯度方向法的适用范围较广。

Negative Gradient Direction Method for Blind Person Exploring the Way

Negative gradient direction method,a common used optimizing method,plays important effect in mechanical engineering field,but,it has the disadvantages of huge calculating amount and low calculating efficiency due to sawtooth phenomenon.One-dimensional blind person pathfingding and optimizing thoughts can be summed as:take three measurements according to 3 situations in relative position of probe point and extreme point.Based on this,improve the negative gradient direction method and give new opti-mizing method-broken line negative gradient direction method.The algorithm is classified into four parts:initial step length inspecting period;inspecting period for double step lengths;the period for temporarily not reducing step length;probing period for half step length.In third part,probe point far end and extreme point are considered.This text offers optimizing thought flow chart and complete C language subprogram. Compared with negative gradient direction method,broken line negative gradient direction method has the advantage of little calculating amount and high optimizing efficiency.Thinking about the situation of stri-ding extreme point,this text gives the algorithm of up one step and back half step.By comparing algorithms of non back half step and non reduced length with back half step,it proves that broken line negative gradi-ent direction method is widely used.