This thesis presents an alternative method for generating a near-optimal closed-loop solution to a non-zero sum perfect information differential game. The alternative method utilizes a concept of fundamental matrix and yields the near-optimal closed-loop solution which is generated by periodically updating the solutions of the two-point boundary-value problem. The resulting updated open-loop control is used between updating intervals.
The advertising differential game model is based on a non-price competition between a new-entering company and its target company. Several numerical examples are solved. The results show that the adaptitive algorithm can be used as an efficient tool to solve the generalized differential game model in which one player can take advantage of the other's non-optimal play.
本 論文은 微分게임에 있어서의 Closed-loop Solution 을 구하는 方法에 關한 것으로써, 特히 本 硏究에서는 從前의 方法이 풀기가 難解하고 適用對象에 따라 不能인 자주 發生되는바 이를 감안하여 풀기가 容易하고 不能인 경우가 거의 發生되지 않는 方法을 誘導하였다. 이를 위해 本 硏究에서는 基本行列槪念(fundamental matrix concept)을 사용하였다.
誘導된 方法을 進入企業과 目標企業間의 廣告費 最適配分競爭問題에 適用시켜서 그 計算結果를 提示하고 旣存의 方法과 比較·檢討 하였다. 앞으로의 硏究方向은 不確實이 介在된 不完全市場下에서도 適用될 수 있는 方法을 開發하는데 注力되어야 할 것이다.