棋類遊戲是人工智慧的分支,隨著電腦科技的進步,電腦在思考邏輯與記憶能力兩方面也有顯著的進展,棋力也隨之提升。利用電腦快速運算的能力和大量記憶的特性,處理棋類遊戲所面臨的難題,以提昇電腦棋類遊戲的棋力,進而教育廣大的棋類遊戲愛好者,成為人工智慧領域中一項有趣的難題。 電腦棋類遊戲的研究,分為中局對局搜尋程式,以及開局與殘局資料庫三部分。我們希望在知識庫的架構上,建立一套由知識擷取、知識分析到知識應用連貫的系統,並廣泛地應用於包括象棋、九路圍棋、黑白棋 … 等各種棋類的開局與殘局上。 本文運用知識庫的技術,在開局部分,用網際網路廣泛蒐集象棋專家的對局資料,建構完成龐大的資料庫。並加以統計分析,形成更有效用的象棋知識庫。在中局部分,有效率地整合開局與殘局的成果。於開局平順地進入中局階段。於中局進入殘局階段,讓電腦將有限的子力導入有利的局勢,再和殘局資料庫銜接,以提升棋力。在殘局部分,以目標導向搜尋的方式,建立實用殘局知識庫。並設計一套有效率的運用方式,讓中局與殘局知識庫、殘局資料庫相結合。 我們以這一套知識庫系統,搭配中局搜尋程式ELP,已取得「中華民國象棋文化協會」六段證書,並於第六、七屆「國際電腦奧林匹亞大賽(Computer Olympiad)」ELP獲得金牌,第八屆獲得銅牌,第九屆結合CONTEMPLATION及ELP分別獲得金牌與銀牌。
Chess is an important field of Artificial Intelligence. As the progress of logic and memory in computer technology, the level of chess program has been promoted. Utilizing computing power and massive memory storage of computer to solve the difficulties in games, to promote the power of computer chess, and to educate numerous amateurs, has become an interesting and difficult problem in Artificial Intelligence. There are three phases in the research in computer chess, including opening, midgame and endgame. Based on the knowledge-base architecture, we built a system from knowledge acquisition, knowledge analysis to knowledge application, and applied it to the opening book and endgame of Chinese Chess, 9*9 Go and Othello. In opening phrase, we built a massive database by collecting game records from Internet, and after statistically analyzed, a more effective Chinese chess knowledge-base was obtained. In midgame phrase, we tried to enter midgame more smoothly after out of book, and let the limited pieces form an advantageous situation so that search engine may combine endgame database in an effective way in endgame phrase. We integrate the knowledge-base system with ELP and CONTEMPLATION, and have awarded the sixth dan certificate of Chinese Chess Culture Society. ELP got the gold metal in the sixth and seventh Computer Olympiad, and the third place in the eighth. In the ninth, CONTEMPLATION and ELP got the first place and the runner-up, respectively.