【摘要】：在Robo Cup2D仿真足球項目中,Agent2D是我國使用最為廣泛的球隊底層之一。仿真平臺中數(shù)據(jù)傳輸?shù)脑肼暩蓴_及代碼自身動作鏈機制不完整等因素,導(dǎo)致采用Agent2D底層的球隊在應(yīng)對不同的隊伍時,存在著適應(yīng)能力不足的缺點,影響了球隊的整體能力。該論文引入了動作修正參數(shù),利用強化學(xué)習(xí)的手段對動作鏈機制進(jìn)行優(yōu)化,使Agent底層球隊在面對不同風(fēng)格的對手時可以選擇更加有效的動作執(zhí)行,以此來提升球隊的適應(yīng)性。仿真實驗證明,此法具有一定效果。
[Abstract]:In the Robo Cup2D soccer simulation project, Agent2D is one of the most widely used teams in our country. The noise interference of data transmission in the simulation platform and the incomplete mechanism of the code itself result in the deficiency of adaptive ability of the teams using Agent2D in dealing with different teams, which affects the overall ability of the team. This paper introduces the motion correction parameters and optimizes the action chain mechanism by means of reinforcement learning so that the Agent team can choose more effective action execution in the face of different styles of opponents in order to improve the adaptability of the team. The simulation results show that this method has certain effect.
【作者單位】： 信息工程大學(xué)理學(xué)院;信息工程大學(xué)指揮軍官基礎(chǔ)教育學(xué)院;安徽工業(yè)大學(xué)計算機學(xué)院;
【分類號】：TP242
，

本文編號：2363109