JP3999282B2 - Computer game device having learning function - Google Patents

Description

更に、上記において、２１ａ−１〜２１ａ−６は、コンピュータ側により制御されるキャラクタ９１の行動を決定する際に参照するデータである。キャラクタ９１と遊戯者により制御されるキャラクタ９０の状態を参照して、いずれのタイプの行動を取るかを決定した後に、この定義テーブルを参照して最適な行動を選択する。
【００５７】
２１ａ−７は、行動パターンを学習する際に参照される。キャラクタのとった行動と、その時に再生されるモーション番号列は、１対１とされているので、ゲーム中に発生したモーション番号列を監視し、ここを参照すればどのような行動をとったかが判る。
【００５８】
更に、２１ａ−８は、実際にコンピュータ側のキャラクタが行動を取る時に参照する。コンピュータ側行動決定プログラム１０２により決まった技を実行する時に、どのようにスイッチをＯＮ／ＯＦＦすればよいかを示している。
【００５９】
次にＲＡＭ３に記憶されるキャラクタ状態データ３１は、ゲームに登場するキャラクタの状態を示している。キャラクタの状態が変化する都度、それに対応して値が変わる。ＣＰＵ１のキャラクタ行動決定手続き１０の処理の結果として、主にこの記憶が更新される。
【００６０】
具体的実施例として、コンピュータ側行動決定プログラム１０２の実行により次にどのような種類の行動を取るべきかを、このキャラクタ状態データ３１（相手が立ち、又はしゃがみの状態か、リング中央か、端か、自分との距離が近いか等の情報）を参照して決定する。
【００６１】
次に、動作結果判定手続き１１の処理によりプレイヤ側及びコンピュータ側のキャラクタ状態データ３１を更新する（キャラクタの表示される座標位置、残りの体力データ等）。尚、キャラクタの表示される座標は２次的なものであり、実際（仮想的）には空間ワールド座標が変化する。これに伴い、表示位置（表示される座標）が変化する。また、キャラクタ状態データ３１は、後に説明するように学習手続き１２の処理のために用いられる。
【００６２】
ＲＡＭ３のプレイヤ個人データ３２の領域に記憶されるデータは、ゲームをしている遊戯者の特徴や能力を行動結果から調べてデータベース化したものである。
【００６３】
具体例としては、キャラクタ状態データ３１のデータから遊戯者がどのような種類の攻撃を出し、またそれが相手に当たったか、防御されたか等の情報を得て、それらをカウントし、遊戯者の行動の傾向を調べる。コンピュータ側行動決定プログラム１０２の実行において、コンピュータ側キャラクタ９１の行動を決定するに際し、このプレイヤ個人データ３２のデータを参照して遊戯者の特徴に合わせてコンピュータ側キャラクタ９１の行動を選択するように制御する。
【００６４】
ついで、ＲＡＭ３に記憶される行動パターンデータ３０は、図６のデータパターンを有する。即ち、この行動パターンデータ３０は、コンピュータ側のキャラクタ９１の現在の行動パターンを示すデータである。例えば、上記ＲＯＭ２の知識データ２１中のどの技をどのようなタイミングで連続して出力するかというようなデータが記憶されている。
【００６５】
ゲーム途中において、ＣＰＵ１のキャラクタ行動決定手続き１０のコンピュータ側行動決定プログラム１０２にしたがいＲＡＭ３に記憶されるキャラクタ状態データ３１、プレイヤ個人データ３２を基に、ＲＯＭ２内の知識データ２１、ＳＲＡＭ４内の学習データ４０のデータにしたがって行動を決定し、行動パターンデータ３０として記憶する。
【００６６】
具体的には、以下のデータ形式で対戦する二人分のデータ領域が確定される。即ち、図６において、３０ｄ−０は、へッダ部であり、キャラクタ状態データ３１、プレイヤ個人データ３２を基にどういう傾向の技を選ぶかという方向性を示すフラグが記録される。
【００６７】
メイン部の３０ｄ─１₀ 〜３０ｄ─１_n は、（ｎ＋１）個の技のそれぞれに付与された番号を示し、３０ｄ─２₀ 〜３０ｄ─２_n は、技と技との間隔即ち、次の技に移行する時間を表示フレーム数で表している。
【００６８】
実施例として、下記のような技に対応して番号が付与される。
【００６９】
３０ｄ−０ フラグ
３０ｄ−１₀ 上段攻撃
３０ｄ−１₁ 下段攻撃
３０ｄ−１₂ 中段攻撃
３０ｄ−１₃ ジャンプ
３０ｄ−１₄ ダウン
３０ｄ−１₅ 立ち投げ
３０ｄ−１₆ しゃがみ投げ
・・・ ・・・
３０ｄ−１₁₅ 攻撃でない
３０ｄ−１₁₆ 大ダメージ
３０ｄ−１₁₇ 早い技
３０ｄ−１₁₈ 射程距離の長い技
３０ｄ−１₁₉ 前方に移動する技
３０ｄ−１₂₀ 後方に移動する技
３０ｄ−１₂₁ 前に移動しない（リングアウトの危険がある場合）
３０ｄ−１₂₂ 後に移動しない（リングアウトの危険がある場合）
更にＲＡＭ３に記憶されるキャラクタ状態遷移データ３３は、学習データを構築するために必要な情報を時系列的に記録している。学習手続き１２において、キャラクタ行動パターン学習プログラム１２２の実行の際に、このキャラクタ状態遷移データ３３のデータを参照して、知識データ２１と照合し、学習データ４０を生成する。
【００７０】
具体例としては、キャラクタがある行動を取る時に、それに対応した動き（モーション）が実行される。したがって、キャラクタ状態データ３１からモーション番号を取り出し、その都度、このキャラクタ状態遷移データ３３の領域に記録する。更に、生成されたモーション番号列を知識データ２１中のモーション番号列２１ａ−７と比較照合させることにより、プレイヤ側キャラクタがどういう行動を取ったかを知ることが出来る。
【００７１】
即ち、図７において、キャラクタ状態遷移データ３３は、第一には、プレイヤ側キャラクタが再生したモーション番号列を意味する。一方、ＣＰＵ側のキャラクタの動き（モーション）も、独立に等価に評価している。したがって、第二にはＣＰＵ側キャラクタのモーション番号例も意味する。
【００７２】
更に、図８は、ＳＲＡＭ４内に記憶される学習データ４０の構成を示す図である。この学習データ４０は、コンピュータ側のキャラクタ行動パターンであり、学習プログラム１２２の実行により生成されるデータである。
【００７３】
具体例として、キャラクタが、ある行動（知識データ２１に記憶されている行動）を取った時に、次にどういう行動を、どういうタイミングで実行するかという情報を記録する。
【００７４】
更に、この学習データ４０のデータ構造は、実施例として次のように構成される。以下のデータ形式で、全キャラクタ分に対し領域が確保される。
【００７５】
４０ａ−０は、ヘッダ部であり記憶している技（最初に出す技）のデータ数ｋ（≦ｎ）を示す。４０ａ−１〜４０ａ−４は、メイン部であり、以下の形式のデータ領域を技毎に最大ｍ個分確保している。
【００７６】
メイン部を個々に説明すると、４０ａ−１は、初めに出す技番号であり、知識データ２１中での番号を示す。４０ａ−２は、次に出す技であり、知識データ２１中での番号で示される。
【００７７】
４０ａ−３は、技の有効度即ち、初めに出す技４０ａ−１と次に出す技４０ａ−２の連携がどれだけ有効かを示す。したがって、実際に技が出された時の結果によって常に変動する。４０ａ−４は、技を出す間隔即ち、最初の技４０ａ−１を出してから次の技４０ａ−２を出すまでの期間を表示フレーム数で示す。
【００７８】
図９は、図８をより理解し易くするための、学習データ４０のデータ構造の概念図である。かかる学習データ４０には、学習手続き１２のプログラム処理により、知識データ２１中のどの技とどの技が連携技として効果的であるかというデータが記憶される。
【００７９】
コンピュータ側行動決定プログラム１０２の実行により、技４０ａ−１の中からその時の状態に適した最初の技４０ａ−１i を選択し、更にそれに続く連携技として技４０ａ−２ijを選択する。
【００８０】
更に、４０ａ−２ij＝４０ａ−１l なる技４０ａ−１のデータが存在する場合は、次の例のように、更に続くデータ４０ａ−２lh を選択できる。
【００８１】
４０ａ−１i ＝ 下パンチ
４０ａ−２ij ＝ 中段キック ＝４０ａ−１l
４０ａ−２lh ＝ 上段キック
次に、上記に説明した、各メモリにおけるデータを参照しながら先に参照した図４の動作フロー、即ちキャラクタ決定手続きＳＳ１の動作を説明する。
【００８２】
コンピュータ側の行動決定プログラム処理（図１の１０２）がスタートすると、先ずキャラクタの状態を示すデータ３１の中からコンピュータ側のキャラクタの行動決定に際し、必要なデータを抽出する（ステップＳ２０）。即ち、ＲＡＭ３のキャラクタ状態データ３１から対戦相手（遊戯者により制御されるキャラクタ９０）とコンピュータ側キャラクタ９１の位置関係や、相手の状況（例えば、立っているのか、しゃがんでいるのか、攻撃中か、体力はどの位残っているか、自己の体力はどの程度かというデータを参照する。
【００８３】
またプレイヤ個人データ３２により、相手がどういう種類の攻撃に弱いかと言う傾向を調査する。このプレイヤ個人データ３２に対する調査もコンピュータ側行動決定プログラム１０２の実行により行われる。
【００８４】
この結果、何らかの行動パターンが行動パターンデータ３０に既にセットされているか否かを判断し（ステップＳ２１）、既にセットされている場合は、他の行動パターンに切り換えた方が有利であるか否かが判断される（ステップＳ２２）。
【００８５】
他の行動パターンに切り換えることが有利と判断される場合は、行動パターンの変更が可能な状態か否かが判断される（ステップＳ２３）。かかる行動パターンの変更が可能な状態か否かの判断は、例えば、技のコマンドの先行入力が可能であり、未だ行動の制御に移行していない場合であっても、既に入力が完了している時は、行動の変更入力は、不可能とされる。
【００８６】
また技を中断することが可能な場合でも技の硬直時間が長い場合等があり、技を変更することは不利となる場合があり、ＲＯＭ２の知識データ２１に基づき、判断される。行動パターンの変更が可能な場合は、知識データ２１及び学習データ４０の中から、現在の状況下で、より有効と思われる行動パターンを選択してセットする。即ち、どのような技が欲しいかを行動パターンデータ３０（図６参照）における、フラグ３０ｄ−０に要求をセットする（ステップＳ２４）。次いで以下のステップにより行動パターンが決定される。
【００８７】
これにより、行動パターンデータ３０に知識データ２１中の技番号が記録されているので、その技番号に対応する技を出すための入力方法（知識データ２１の２１ａ−８：図５参照）に従いスイッチを入力する、即ちセットされているパターンデータにしたがい、キャラクタの行動が決定される（ステップＳ２５）。この結果、パターン選択は終了し（ステップＳ２６）、所定の終了コード（例えば０×７ｆ）を行動パターンデータ３０にセットする（ステップＳ２７）。
【００８８】
更に、上記ステップＳ２４の詳細な処理の内容は、サブルーチンとして図１０、図１１に示される。図１０及び図１１は、一連のフローであり、分割して図１０にステップＳ２４の詳細（その１）、図１１にステップＳ２４の詳細（その２）として示している。
【００８９】
先ず図１０において、フラグ３０ｄ−０をセットしてコールし、行動パターンデータ３０を初期化する。この初期化において、３０ｄ─１は、終了コード（ｅｎｄ−ｃｏｄｅ：０×７ｆ）で、３０ｄ−２は、初期コード例えば、０でそれぞれを埋める。
【００９０】
ついで、学習データ４０が存在するか否かを判断し（ステップＳ３１）、存在する場合は、学習データ４０（図８参照）中の４０ａ−１の中に行動パターンデータ３０のフラグ３０ｄ−０を満たす技があるか否かを判断する（ステップＳ３２）。
【００９１】
フラグ３０ｄ−０を満たす技があれば、学習データ４０中の４０ａ−１の中で行動パターンデータ３０のフラグ３０ｄ−０を満たす技を全て、バッファに取り出し、その中から任意の技４０ａ−１i を選び、３０ｄ−１にセットする（ステップＳ３３）。
【００９２】
一方、ステップＳ３２において、フラグ３０ｄ−０を満たす技がない場合は、学習データ４０中の４０ａ−１から任意の技４０ａ−１i を選び、行動パターンデータ３０の３０ｄ−１にセットする（ステップＳ３４）。
【００９３】
次に、図１１の処理フローに継続し、学習データ４０の技４０ａ−１i から続く技４０ａ−２i のうちフラグ３０ｄ−０を満たす技があるか否かを判断する（ステップＳ３５）。フラグ３０ｄ−０を満たす技がある場合は、学習データ４０の技４０ａ−２i の中でフラグ３０ｄ−０を満たす技を全てバッファに取り出し、その中から任意の枝４０ａ−２ijを選び、技３０ｄ−１にセットする（ステップＳ３６）。
【００９４】
一方、ステップＳ３５において、フラグ３０ｄ−０を満たす技が無い場合は、技４０−２i の中から任意の技４０−２ijを選び、技３０ｄ−１にセットする（ステップＳ３７）。
【００９５】
ステップＳ３５及び３６に続き、技の間隔４０ａ−４ijを３０ｄ−２セットする（ステップＳ３８）。次いで、３０ｄ−１の領域全てにデータをセットしたか否かを判断する（ステップＳ３９）。３０ｄ−１の領域全てにデータをセットしていなければ、４０ａ−１中に４０ａ−２ij＝４０ａ−１l なる技が存在するか否かを判断する（ステップＳ４０）。ステップＳ４０でＹＥＳの場合は、ステップＳ３５に戻る。
【００９６】
更に、ステップＳ４０でＮＯの場合及び、ステップＳ３９でＹＥＳの場合、１度もフラグ３０ｄ−０を満たす技を選ばなかったかを判断し（ステップＳ４１）、そうであれば、行動パターンデータ３０ｄ−１、３０ｄ−２ａを初期化する（ステップＳ４２）。
【００９７】
ついで、ステップＳ４２の後、及びステップＳ３１でＮＯの場合、知識データ２１の中からフラグ３０ｄ−０を満たす技の全てをバッファに取り出し、その中から任意の技を選び３０ｄ−１にセットする（ステップＳ４３）。ステップＳ４１及びステップＳ４３の後は、上記の図１０及び図１１の処理が繰り返される。
【００９８】
図１２は、図３における学習手続きＳＳ２の詳細フローを示す図である。先ずキャラクタの状態を示すデータ３１の中から行動パターンを生成するのに必要なデータを抽出する（ステップＳ５０）。この処理は、モーション番号をキャラクタ状態データ３１から取り出して、キャラクタ状態遷移データ３３にモーション番号列として記録して行う。
【００９９】
次に十分なデータが集まり、パターンを生成可能かを判断する（ステップＳ５１）。これは、モーション番号列と知識データ２１との一致を調べるのに十分なモーション番号列が得られたか否かで判断する。
【０１００】
パターンの生成が可能でなければ終了し、パターンを生成可能であれば、状態遷移の仕方を基に行動パターンを生成する（ステップＳ５２）。これは、キャラクタ状態遷移データ３３と知識データ２１の一致を行い２つの技番号を得ることにより行う。尚、上記ステップＳ５０〜Ｓ５２の内容は、適用するゲームの内容により異なってくる。
【０１０１】
この生成された行動パターンをもとに学習が行われる（ステップＳ５３）。この行動パターンの学習（ステップＳ５３）の内容は、図１３に示される。図１３のフローにおいて、まず、既に学習済の行動パターンかを判断する（ステップＳ５３１）。既に学習済の行動パターンでなければ、記憶領域に空きが有るか否かを判断する（ステップＳ５３２）。空きがなければ、優先度の低いパターンを消去し（ステップＳ５３３）、新しい行動パターンを学習する（ステップＳ５３４）。この消去は、学習データ４０中の４０ａ−３に相当する値の低いものを消去して行う。
【０１０２】
空きが有ればそのまま新しい行動パターンを学習する（ステップＳ５３４）。行動パターンの学習は、学習データ４０に記録される。
【０１０３】
更に、既に学習済の行動パターンである場合（ステップＳ５３１において、ＹＥＳの場合）、及び新しい行動パターンを学習した場合（ステップＳ５３４）、ついで勝敗結果に応じて学習した行動パターンの優先度を変化させる（ステップＳ５３５）。即ち、行動パターンの実行結果（技が相手に当たったか否か）によって、学習データ４０中の４０ａ−３に相当する値を変化させる。
【０１０４】
更に、図１４は、更に学習手続きＳＳ２の実施例詳細フローを示す図であり、実施例行動パターンとして連携技を学習する場合の動作フローである。図１２との関連において、図１４の動作を説明する。
【０１０５】
図１４において、ステップＳ６０は、図１２のステップＳ５０に対応し、現在実行中のモーション番号を取り出して、キャラクタ状態遷移データ３３中のエンドモーションポインタが示すメモリＭ_mに記録し、モーション番号列３３ｃ−０を生成する（ステップＳ６０）。この時、エンドモーションポインタは、次のＭ_m+1を指す。Ｍ_mがバッファエンドの場合は、エンドモーションポインタＡはバッファ先頭を指す。
【０１０６】
次に知識データ２１中に登録してある各技のモーション番号列２１ａ─７ｊと生成したモーション番号列３３ｃ−０を比較して現在キャラクタがどのような行動をとっているかを調べる（ステップＳ６１）。
【０１０７】
その後、３３ｃ−１に既に認識された技が入っているかを判断する（ステップＳ６２）。ＮＯの場合は、３３ｃ−０と２１ａ−７のモーション番号列のマッチングがとれ、技が確定したか否かを判断し（ステップＳ６３）、確定していなければ終了し、確定していれば３３ｃ−１に技番号（知識データ２１に登録されている番号）を記録する（ステップＳ６４）。
【０１０８】
一方、ステップＳ６２において、ＹＥＳの場合は、３３ｃ−０ｊと２１ａ−７のモーション番号列のマッチングがとれ、技が確定したかを判断する（ステップＳ６５）。技が確定していなければ、３３ｃ−３のタイマー（技と技の間隔）を１だけ増加（ステップＳ６６）して終了する。
【０１０９】
技が確定していれば、３３ｃ−３のタイマー（技と技の間隔）が閾値より小さいかを判断し（ステップＳ６７）、小さい場合は、３３ｃ−１、３３ｃ−２及び３３ｃ−３を行動パターンとして、ＳＲＡＭ４に学習データ４０として記録する（ステップＳ６８）。
【０１１０】
このステップＳ６８の内容は、図１２の行動パターンを学習するステップＳ５３に対応し、従ってその詳細は先に説明した図１３のとおりである。
【０１１１】
ステップＳ６７において、３３ｃ−３のタイマー（技と技の間隔）が閾値より小さくない時及び、ステップＳ６８の学習データ４０として記録が行われた時、３３ｃ−３ｋ タイマー（技と技の間隔）をクリヤする、したがって、３３ｃ−３＝空列（ヌル）とする（ステップＳ６９）。ついで、３３ｃ−１に３３ｃ−２を入れ、３３ｃ−２をクリアして終了する。
【０１１２】
【発明の効果】
以上実施例にしたがい説明したように、本発明により遊戯者がゲームをする度にコンピュータ側の反応が変化する。これにより、遊戯者の上達度の向上に伴い、コンピュータ側もより高度で有効な戦術を用いるようになり遊戯者の興味を維持し続けることが可能である。
【０１１３】
また、遊戯者の特徴、技能を判断し、行動に影響を与えることにより遊戯者に応じた難易度の調整ができるために初心者でも取組やすい。更に、上記実施例は本発明の実施例であり、本発明はこれらの実施例に限定されるものではない。また、本発明は、コンピュータと遊戯者とで対決する形式のゲーム全てに適用可能である。更に、従来のように新しいゲームを制作する際に、ゲーム中起こり得るあらゆる状況を想定してコンピュータ側の行動パターンを作成する手間が大幅に減少される。
【図面の簡単な説明】
【図１】本発明の実施例装置のブロック図である。
【図２】モニタ表示画面の一例である。
【図３】本発明ゲーム装置の全体動作フローを示す図である。
【図４】図３のキャラクタ行動決定手続きＳＳ１の詳細フローを示す図である。
【図５】知識データを説明する図である。
【図６】行動パターンデータを説明する図である。
【図７】キャラクタ状態遷移データを説明する図である。
【図８】学習データを説明する図である。
【図９】図８の学習データの組立てを説明する図である。
【図１０】図４のステップＳ２４の詳細（その１）を説明するフローである。
【図１１】図４のステップＳ２４の詳細（その２）を説明するフローである。
【図１２】図３の学習手続きＳＳ２の詳細フローを示す図である。
【図１３】行動パターンの学習フローを示す図である。
【図１４】行動パターンを連携技とする学習手続きの実施例フローを示す図である。
【符号の説明】
１ ＣＰＵ本体
２、５ＲＯＭ
３ ＲＡＭ
４ ＳＲＡＭ
６ 入力装置
７ 出力制御回路
８ 出力装置
１０ キャラクタ行動決定手続き
１１ 動作結果判定手続き
１２ 学習手続き
１３ 出力制御手続き
２０ ゲームプログラム本体
２１ 知識データ
３０ 行動パリーンデータ
３１ キャラクタ状態データ
３２ プレイヤ個人データ
３３ キャラクタの状態遷移データ
４０ 学習データ
４１ 出力データ[0001]
[Industrial application fields]
The present invention relates to a computer game device, and in particular, in a battle game between a character controlled by a player and a character controlled from the computer side, the movement of the character by the control from the computer side is uniform with a predetermined content (one pattern) The present invention relates to a computer game device having a learning function.
[0002]
[Prior art]
In recent years, with the high integration and high speed of microcomputers, computer game devices such as video game devices have become widespread and their contents have become more complicated.
[0003]
The spread of such computer game devices depends on the player's interest based on the contents of the game program. In particular, when the behavior of the character controlled by the computer becomes uniform with respect to the movement of the character controlled by the player, the content of the game becomes uninteresting.
[0004]
That is, in a conventional computer game device, when a game device is produced or a game program is created, when several behavior patterns of the character and various developments during game execution are assumed and a specific condition is satisfied A pattern selection method for selecting an action to be selected is set in advance in a program, a database, or the like.
[0005]
With such conventional technology, the difficulty level and development of the game can be easily set as intended by the producer. In such a case, it is easy to establish a game, but the development of the game and the execution contents of the computer tend to be uniform (one pattern).
[0006]
For this reason, the player is prefetched about the execution contents of the computer, and the game itself is bored. In addition, when a player's tactics or game situation that was not predicted at the time of production or that could not be handled due to a production error is discovered, the computer can almost always cope with these problems. It ends without.
[0007]
As described above, when the game contents (tactics) can be performed by simple operations, the player's interest in the game is immediately halved and the life of the game program is shortened.
[0008]
In order to make it difficult for a player to execute such a certain work tactic, in the conventional apparatus, a method for increasing the reaction speed of the character or stricter the condition on the player side has been tried. However, in this method, there is a problem that the difficulty of the game is increased in a mischievous manner, and if the player is a beginner, the game itself loses interest.
[0009]
[Problems to be solved by the invention]
Accordingly, an object of the present invention is to provide a computer game device having a learning function for solving the problems in the above-described conventional devices.
[0010]
Specifically, an object of the present invention is to automatically learn a character's action pattern under computer control on the computer side when the game is executed, and after learning, the action pattern of the character controlled by the computer. An object of the present invention is to provide a computer game device having a learning function that can be executed.
[0011]
[Means for Solving the Problems]
A computer game device having a learning function according to the present invention includes an input device to which data for controlling an action of a first character is input by a player's operation, and an action of a second character that plays against the first character. A CPU for controlling, an action that a character appearing in the game can take, a memory for storing the result of the action as knowledge data, and an output device for displaying the game under the control of the CPU. A database for determining the action of the second character to be taken next time is constructed based on the knowledge data stored in the memory with reference to the state of the character appearing in the game based on the result of .
[0012]
Further, as a specific configuration, in the computer game device having the learning function of the present invention, the memory further obtains the characteristics and game ability of the player who plays the game from the action result of the first character, and creates a database. Character state transition data constituted by arranging the performed player personal data and the executed actions of the first and second characters in time series are stored. And CPU is comprised so that the database for determining the action after the next time of said 2nd character based on these data may be constructed | assembled.
[0013]
In addition, a second memory for storing a database for determining the next and subsequent actions of the second character as learning data is further provided, and the learning data stored in the second memory is combined with the knowledge data. The second character is configured to be used for determining the next and subsequent actions.
[0014]
[Action]
According to the present invention, the computer-side character action pattern is based on the knowledge data registered in advance during the game and the state transition state of the character resulting from the player-side and computer-side action during the game. Created and stored as learning data as needed.
[0015]
This learned and stored action pattern is referred to as one of the subsequent computer-side character action patterns, equivalent to the previously registered knowledge data. As described above, in the present invention, the action pattern of the computer-side character is newly accumulated by learning during the game and used for the subsequent games.
[0016]
Therefore, it is possible to avoid that the action pattern of the computer character in the conventional game device is uniform.
[0017]
【Example】
FIG. 1 is a block diagram of an embodiment of a computer game device according to the present invention. In the figure, reference numeral 1 denotes a control unit of the game apparatus main body, which is a CPU main body. The CPU main body 1 controls the entire game device described later based on the game program main body 20 stored in the ROM 2.
[0018]
Further, the CPU body 1 similarly performs a plurality of subroutine processes to execute the features of the present invention based on the control of the game program body 20. These subroutine processes are shown as blocks in the CPU body 1 as a character action determination procedure 10, an action result determination procedure 11, a learning procedure 12, and an output control procedure 13. The details of each procedure will be described later.
[0019]
The CPU main body 1 is configured to include each memory of ROM 2, 5, RAM 3, SRAM 4, and an output device 8 such as a display monitor connected through the input device 6 and the output control circuit 7.
[0020]
The detailed configuration and operation of the apparatus of the present invention will be described below for a game program that executes a fighting game as shown in FIG. FIG. 2 shows an example of the monitor screen of the output device 8. That is, in FIG. 2, 9 is a monitor screen, and 90 and 91 are fighting characters.
[0021]
90 is a character whose movement is controlled by an operation of the player's input device 6, and 91 is a character whose movement is controlled by predetermined program control on the computer side.
[0022]
Reference numeral 92 denotes an area for displaying a game time limit. Depending on the display method, the time limit of the game is displayed as the elapsed time of the game or the time until the end of the game. Reference numerals 93 and 94 denote areas for displaying the physical strength gauges of the characters 90 and 91, respectively.
[0023]
The displayed martial arts game is a fighting martial art game in which a character plays a technique such as punching and kicking one-on-one and fights. When the attack of one character hits the opponent character, the values of the physical strength gauges 93 and 94 of the opponent character are reduced accordingly.
[0024]
If the opponent's physical strength gauge is set to 0 within the time limit displayed in the area 92 or the opponent is pushed out of the ring 95, the game is won. Further, when the characters 90 and 91 are in the ring 95 and the time limit has passed, the character with the remaining physical strength gauge wins.
[0025]
Returning to FIG. 1, the ROM 2 stores knowledge data 21 in addition to the game program. Here, the knowledge data are actions that can be taken by each character appearing in the game and the corresponding results. For example, all the actions that each character can take according to the data input by the input device 6 are stored as data.
[0026]
The RAM 3 is a work RAM, and stores action pattern data 30, character state data 31, player personal data 32, and character state transition data 33. Furthermore, the SRAM 4 stores learning data 40. The ROM 5 stores output data 50 such as image data and audio data.
[0027]
Therefore, the player can act in the game by operating the input device 6 based on the information displayed on the output device 8. That is, data input from the input device 6 is sent to the CPU body 1 and processed by the player action determination program 101.
[0028]
At the same time, the CPU body 1 is processed by the computer-side action determination program 102 with reference to the knowledge data 21 stored in the ROM 1, the character state data 31 stored in the RAM 3, the player personal data 32, and the learning data 40 of the SRAM 4. Then, the action of the computer character is determined.
[0029]
If the actions of the characters on the player side and the computer side are determined in this way, the state of each character is changed based on this. In response to this change in state, the game result determination procedure 11 determines whether the game is temporary or final. Based on the determination result, the character state data 31 of the RAM 3 is updated and stored.
[0030]
Further, the CPU body 1 reads out necessary data from the character state data 31 of the RAM 3 based on the processing of the learning procedure 12 and stores it in the RAM 3 as the character transition state data 33.
[0031]
The CPU body 1 refers to necessary data in the character state data 31 in the RAM 3, and stores the player personal data 32 in the RAM 3 in accordance with the player characteristics in the process of the learning procedure 12 and the process of the ability determination program 121. Create and update.
[0032]
Furthermore, the CPU body 1 refers to the character state transition data 33 in the RAM 3 and creates an action pattern according to the processing of the computer-side character action pattern learning program 122. Then, the created action pattern is updated and stored in the SRAM 4 as learning data 40.
[0033]
Furthermore, the operation of the game device according to the present invention as an overall configuration is controlled based on the overall game flow of FIG. The operation based on this flow is executed under the control of the CPU 1 based on the game program main body 20 stored in the ROM 2.
[0034]
3, when the power is turned on or reset by a switch not shown in FIG. 1 (step S1), the work RAM 3 is initialized (step S2). Next, it is determined whether or not to start the game (step S3).
[0035]
When starting the game, the character action determination procedure flow SS1 is executed. In the character action determination procedure flow SS1, it is first determined that a key on the player (player) side is input (input is performed by the input device 6) (step S4).
[0036]
Next, the next action of the character 91 controlled by the computer is considered (step S5). Based on this, key input on the computer side is performed (step S6). Then, the action of the character 91 on the computer side is determined and executed in response to the key input on the computer side (step S7). Here, key input on the computer side means electronic processing equivalent to switch input by a program.
[0037]
Next, the operation result is determined (step S8). Here, in the player side input determination (step S4) and the player side character motion determination (step S7), the CPU 1 detects an input from the input device 6 and performs the subroutine of the character action determination procedure 10 which is a subroutine process. Of these, the player side action determination program is executed (step S101).
[0038]
On the other hand, the thinking of the next action of the character 91 controlled by the computer side (step S5), the determination of key input on the computer side (step S6), and the action determination of the character on the computer side (step S7) are performed in a subroutine process in the CPU1. Of the certain character action determination procedure 10, it is executed according to the computer-side action determination program 102.
[0039]
In the flow of FIG. 3, the result of both character actions is then determined (step S8). This determination is executed by the operation result determination procedure 11 which is a subroutine process in the CPU 1.
[0040]
Thereafter, the learning procedure SS2 is executed. The learning procedure SS2 is executed by the learning procedure 12 which is a subroutine process in the CPU 1 of FIG. Further, in the learning procedure 12, the player characteristics and ability are determined based on the action result of the player character 90 by the player characteristics and ability determination program 121, and recorded as personal data in the player personal data area 32 of the RAM 3. (Step S9).
[0041]
Further, the character action pattern learning program 122 is executed in the learning procedure 12, which is a subroutine process in the CPU 1, and the result is stored in the SRAM 4 as learning data 40 (step S10).
[0042]
Next, in the flow of FIG. 3, according to the progress of the game, the image and sound data 50 stored in the ROM 5 are output to the output device 8 through the output control circuit 7 under the output control procedure 13 of the CPU 1 (step S11). .
[0043]
Further, it is determined whether or not the game is over (step S12). If the game is over, the process returns to step S3.
[0044]
Next, the character action determination procedure SS1 and the learning procedure SS2 executed according to the present invention in the flow of FIG. 3 will be sequentially described in detail.
[0045]
FIG. 4 is a flow for explaining the operation by the program for determining the action pattern on the computer side of the character action determination procedure 10 in the CPU 1 of FIG. In describing this flow, in order to better understand the characteristics of the present invention, the contents of the knowledge data 21 in the ROM 2, the data 30 to 33 in the RAM 3, and the learning data 40 in the SRAM 4 will be described first.
[0046]
FIG. 5 shows the contents of an area for storing the knowledge data 21 in the ROM 2. It is displayed in a data format as shown in FIG. 5, and data for all characters is prepared. That is, in FIG. 5, 21a-0 is a header part, which indicates the number of registered techniques (maximum m).
[0047]
Now, assuming that the number of registered techniques is k, the contents of the technique are defined by 21a-1 to 21a-8 for each technique in the main part. For example, 21a-1 _k-1 ~ 21a-8 _k-1 Is the definition of the k-th technique.
[0048]
As an example, the contents of each definition will be described as follows. 21a-1 shows the attribute of a technique and whether it is an attack. Define what kind of attack it is.
[0049]
21a-2 indicates the damage of the technique. If it is an attack, its power is set. If not, 0 is set.
[0050]
21a-3 indicates the speed of the attack, and is indicated by the number of display frames at which the attack determination starts to appear. 0 for non-throwing or attacking techniques.
[0051]
Reference numeral 21a-4 denotes a rigid time, which is indicated by the number of display frames from when the operation is executed until the next action can be started.
[0052]
21a-5 indicates the range of the attack, reach distance of the attack, throw indicates the throwing time, and indicates 0 when not attacking.
[0053]
21a-6 shows a movement distance and shows how much it moves back and forth after the end of the action.
[0054]
21a-7 shows a column (maximum of 5) of motion numbers to be reproduced when taking action. Note that the action of the character means reproducing a motion corresponding to the action.
[0055]
21a-8 is an input command, and defines an input operation procedure of the input device 6 necessary for taking the action of the technique defined here. For example, it is defined in the following procedure.
[0056]

Further, in the above, 21a-1 to 21a-6 are data to be referred to when determining the action of the character 91 controlled by the computer side. After deciding which type of action to take with reference to the state of the character 91 and the character 90 controlled by the player, the optimum action is selected with reference to this definition table.
[0057]
21a-7 is referred to when learning a behavior pattern. The action taken by the character and the motion number sequence that is played at that time are one-to-one. I understand.
[0058]
Furthermore, 21a-8 is referred to when the computer character actually takes action. It shows how the switch should be turned ON / OFF when a technique determined by the computer-side action determination program 102 is executed.
[0059]
Next, the character state data 31 stored in the RAM 3 indicates the state of the character appearing in the game. Each time the character's state changes, the value changes accordingly. This memory is mainly updated as a result of the processing of the character action determination procedure 10 of the CPU 1.
[0060]
As a specific example, the character state data 31 (whether the opponent is standing or crouching, the center of the ring, the end, Or information on whether or not the user is close to the user).
[0061]
Next, the player-side and computer-side character state data 31 is updated by the processing of the motion result determination procedure 11 (coordinate positions where characters are displayed, remaining physical strength data, etc.). Note that the coordinates displayed by the character are secondary, and the space world coordinates actually change (virtually). Accordingly, the display position (displayed coordinates) changes. The character state data 31 is used for the processing of the learning procedure 12 as will be described later.
[0062]
The data stored in the area of the player personal data 32 in the RAM 3 is a database obtained by examining the characteristics and abilities of the player who is playing the game from the action results.
[0063]
As a specific example, from the data of the character state data 31, information such as what kind of attack the player has made, how it hit the opponent, whether it was defended, etc. is obtained, counted, and the player's Investigate behavioral trends. In the execution of the computer-side action determination program 102, when determining the action of the computer-side character 91, the action of the computer-side character 91 is selected according to the player's characteristics with reference to the data of the player personal data 32. Control.
[0064]
Next, the action pattern data 30 stored in the RAM 3 has the data pattern of FIG. That is, the action pattern data 30 is data indicating the current action pattern of the character 91 on the computer side. For example, data such as which technique in the knowledge data 21 of the ROM 2 is continuously output at what timing is stored.
[0065]
In the middle of the game, based on the character state data 31 and player personal data 32 stored in the RAM 3 in accordance with the computer-side action determination program 102 of the character action determination procedure 10 of the CPU 1, knowledge data 21 in the ROM 2 and learning data in the SRAM 4. The behavior is determined according to the 40 data and stored as behavior pattern data 30.
[0066]
Specifically, the data area for two players who will compete in the following data format is determined. That is, in FIG. 6, 30 d-0 is a header portion, and a flag indicating the direction of the skill to be selected based on the character state data 31 and the player personal data 32 is recorded.
[0067]
30d-1 of main part ₀ ~ 30d-1 _n Indicates the number assigned to each of (n + 1) techniques, and 30d-2 ₀ ~ 30d-2 _n Represents the interval between techniques, that is, the time to shift to the next technique, by the number of display frames.
[0068]
As an example, numbers are assigned corresponding to the following techniques.
[0069]
30d-0 flag
30d-1 ₀ Upper attack
30d-1 ₁ Lower attack
30d-1 ₂ Middle attack
30d-1 _Three Jump
30d-1 _Four down
30d-1 _Five Standing
30d-1 ₆ Crouch throw
...
30d-1 ₁₅ Not attack
30d-1 ₁₆ Heavy damage
30d-1 ₁₇ Quick trick
30d-1 ₁₈ Long range
30d-1 ₁₉ Move forward
30d-1 ₂₀ Technique to move backward
30d-1 _{twenty one} Do not move forward (when there is a risk of ringout)
30d-1 _{twenty two} Do not move after (when there is a risk of ringout)
Further, the character state transition data 33 stored in the RAM 3 records information necessary for constructing learning data in time series. In the learning procedure 12, when the character behavior pattern learning program 122 is executed, the character state transition data 33 is referred to and collated with the knowledge data 21 to generate learning data 40.
[0070]
As a specific example, when a character takes an action, a motion corresponding to the action is executed. Therefore, the motion number is extracted from the character state data 31 and recorded in the character state transition data 33 area each time. Furthermore, by comparing the generated motion number sequence with the motion number sequence 21a-7 in the knowledge data 21, it is possible to know what action the player-side character has taken.
[0071]
That is, in FIG. 7, the character state transition data 33 first means a motion number sequence reproduced by the player-side character. On the other hand, the movement (motion) of the character on the CPU side is also evaluated independently and equivalently. Therefore, the second means an example of the motion number of the CPU character.
[0072]
Further, FIG. 8 is a diagram showing a configuration of learning data 40 stored in the SRAM 4. The learning data 40 is a character action pattern on the computer side, and is data generated by executing the learning program 122.
[0073]
As a specific example, when a character takes a certain action (the action stored in the knowledge data 21), information on what kind of action is executed next and at what timing is recorded.
[0074]
Furthermore, the data structure of the learning data 40 is configured as follows as an example. Areas are reserved for all characters in the following data format.
[0075]
40a-0 is a header part and indicates the number of data k (≦ n) of the stored technique (first technique to be issued). Reference numerals 40a-1 to 40a-4 denote main parts, which secure a maximum of m data areas of the following format for each technique.
[0076]
The main part will be described individually. 40a-1 is a technique number to be issued first, and indicates a number in the knowledge data 21. 40a-2 is a technique to be issued next, and is indicated by a number in the knowledge data 21.
[0077]
40a-3 indicates the effectiveness of the technique, that is, how effective the cooperation between the technique 40a-1 to be issued first and the technique 40a-2 to be issued next is effective. Therefore, it always fluctuates depending on the result when the technique is actually put out. Reference numeral 40a-4 indicates a time interval for giving a technique, that is, a period from the first technique 40a-1 to the next technique 40a-2 in terms of the number of display frames.
[0078]
FIG. 9 is a conceptual diagram of the data structure of the learning data 40 for easier understanding of FIG. The learning data 40 stores data indicating which techniques in the knowledge data 21 and which techniques are effective as cooperative techniques by the program processing of the learning procedure 12.
[0079]
By executing the computer-side action determination program 102, the first technique 40a-1i suitable for the current state is selected from the techniques 40a-1, and the technique 40a-2ij is selected as the subsequent cooperative technique.
[0080]
Further, when there is data of technique 40a-1 such that 40a-2ij = 40a-1l, the following data 40a-2lh can be selected as in the following example.
[0081]
40a-1i = lower punch
40a-2ij = middle kick = 40a-1l
40a-2lh = upper kick
Next, the operation flow of FIG. 4 referred to above, that is, the operation of the character determination procedure SS1 will be described with reference to the data in each memory described above.
[0082]
When the computer-side action determination program process (102 in FIG. 1) starts, first, necessary data is extracted from the data 31 indicating the character state when determining the action of the computer-side character (step S20). That is, the positional relationship between the opponent (character 90 controlled by the player) and the computer-side character 91 from the character state data 31 in the RAM 3 and the situation of the opponent (for example, standing, crouching, or attacking) Refer to the data on how much physical strength remains and how much is your physical strength.
[0083]
The player personal data 32 is used to investigate the tendency of the opponent to be vulnerable to what type of attack. The survey of the player personal data 32 is also performed by executing the computer side action determination program 102.
[0084]
As a result, it is determined whether or not any action pattern is already set in the action pattern data 30 (step S21). If it is already set, whether or not it is advantageous to switch to another action pattern. Is determined (step S22).
[0085]
When it is determined that switching to another behavior pattern is advantageous, it is determined whether or not the behavior pattern can be changed (step S23). For example, it is possible to determine whether or not the behavior pattern can be changed by, for example, prior input of a technique command, even if the transition to behavior control has not yet been completed. When it is, the change input of the action is impossible.
[0086]
Even if the technique can be interrupted, there are cases in which the technique has a long rigid time, and it may be disadvantageous to change the technique, and the determination is made based on the knowledge data 21 in the ROM 2. When the behavior pattern can be changed, the behavior pattern that is considered to be more effective under the current situation is selected and set from the knowledge data 21 and the learning data 40. That is, a request is set in the flag 30d-0 in the action pattern data 30 (see FIG. 6) to determine what technique is desired (step S24). Next, an action pattern is determined by the following steps.
[0087]
Thereby, since the technique number in the knowledge data 21 is recorded in the action pattern data 30, the switch is performed according to the input method (21a-8 of the knowledge data 21: see FIG. 5) for taking out the technique corresponding to the technique number. Is input, that is, the action of the character is determined in accordance with the set pattern data (step S25). As a result, the pattern selection ends (step S26), and a predetermined end code (for example, 0 × 7f) is set in the action pattern data 30 (step S27).
[0088]
Further, details of the processing in step S24 are shown as subroutines in FIGS. 10 and 11 show a series of flows, which are divided and shown in FIG. 10 as details of step S24 (part 1) and FIG. 11 as details of step S24 (part 2).
[0089]
First, in FIG. 10, the flag 30d-0 is set and called to initialize the action pattern data 30. In this initialization, 30d-1 is filled with an end code (end-code: 0 × 7f), and 30d-2 is filled with an initial code, for example, 0.
[0090]
Next, it is determined whether or not the learning data 40 exists (step S31). If it exists, the flag 30d-0 of the action pattern data 30 is set in 40a-1 in the learning data 40 (see FIG. 8). It is determined whether there is a technique to satisfy (step S32).
[0091]
If there is a technique satisfying the flag 30d-0, all the techniques satisfying the flag 30d-0 of the action pattern data 30 among the 40a-1 in the learning data 40 are taken out to the buffer, and any technique 40a-1i is selected from them. Is selected and set to 30d-1 (step S33).
[0092]
On the other hand, if there is no technique that satisfies the flag 30d-0 in step S32, an arbitrary technique 40a-1i is selected from 40a-1 in the learning data 40 and set to 30d-1 in the action pattern data 30 (step S34). ).
[0093]
Next, continuing from the processing flow of FIG. 11, it is determined whether or not there is a technique satisfying the flag 30d-0 among the techniques 40a-2i following the technique 40a-1i of the learning data 40 (step S35). If there is a technique satisfying the flag 30d-0, all the techniques satisfying the flag 30d-0 among the techniques 40a-2i of the learning data 40 are taken out to the buffer, and an arbitrary branch 40a-2ij is selected from them, and the technique 30d is selected. Set to -1 (step S36).
[0094]
On the other hand, if there is no technique that satisfies the flag 30d-0 in step S35, an arbitrary technique 40-2ij is selected from the techniques 40-2i and set to the technique 30d-1 (step S37).
[0095]
Subsequent to steps S35 and S36, the technique interval 40a-4ij is set to 30d-2 (step S38). Next, it is determined whether or not data has been set in all the areas 30d-1 (step S39). If the data is not set in all the areas 30d-1, it is determined whether or not a technique of 40a-2ij = 40a-1l exists in 40a-1 (step S40). If YES in step S40, the process returns to step S35.
[0096]
Further, if NO in step S40 and YES in step S39, it is determined whether a technique that satisfies the flag 30d-0 has been selected (step S41). If so, action pattern data 30d-1 , 30d-2a are initialized (step S42).
[0097]
Then, after step S42 and in the case of NO at step S31, all the techniques satisfying the flag 30d-0 are extracted from the knowledge data 21 to the buffer, and an arbitrary technique is selected from them and set to 30d-1 ( Step S43). After step S41 and step S43, the processes of FIGS. 10 and 11 are repeated.
[0098]
FIG. 12 is a diagram showing a detailed flow of the learning procedure SS2 in FIG. First, data necessary for generating an action pattern is extracted from the data 31 indicating the character state (step S50). This process is performed by extracting the motion number from the character state data 31 and recording it in the character state transition data 33 as a motion number string.
[0099]
Next, it is determined whether sufficient data is collected and a pattern can be generated (step S51). This determination is made based on whether or not a sufficient motion number sequence has been obtained for checking the match between the motion number sequence and the knowledge data 21.
[0100]
If the pattern cannot be generated, the process ends. If the pattern can be generated, an action pattern is generated based on the state transition method (step S52). This is done by matching the character state transition data 33 and the knowledge data 21 to obtain two technique numbers. Note that the contents of steps S50 to S52 differ depending on the contents of the game to be applied.
[0101]
Learning is performed based on the generated behavior pattern (step S53). The contents of this behavior pattern learning (step S53) are shown in FIG. In the flow of FIG. 13, it is first determined whether or not the action pattern has already been learned (step S531). If it is not an already learned action pattern, it is determined whether or not there is a free space in the storage area (step S532). If there is no space, the low priority pattern is deleted (step S533), and a new behavior pattern is learned (step S534). This erasure is performed by erasing a low value corresponding to 40a-3 in the learning data 40.
[0102]
If there is a vacancy, a new action pattern is learned as it is (step S534). The learning of the behavior pattern is recorded in the learning data 40.
[0103]
Furthermore, when the action pattern has already been learned (in the case of YES in step S531) and when a new action pattern has been learned (step S534), the priority of the learned action pattern is changed according to the winning / losing result. (Step S535). That is, the value corresponding to 40a-3 in the learning data 40 is changed according to the action pattern execution result (whether or not the technique hits the opponent).
[0104]
Further, FIG. 14 is a diagram showing an example detailed flow of the learning procedure SS2, and is an operation flow in the case of learning a cooperative technique as an example action pattern. The operation of FIG. 14 will be described in relation to FIG.
[0105]
In FIG. 14, step S60 corresponds to step S50 of FIG. 12, takes out the currently executed motion number, and indicates the memory M indicated by the end motion pointer in the character state transition data 33. _m And the motion number string 33c-0 is generated (step S60). At this time, the end motion pointer is moved to the next M _{m + 1} Point to. M _m Is a buffer end, the end motion pointer A points to the head of the buffer.
[0106]
Next, the motion number sequence 21a-7j of each technique registered in the knowledge data 21 is compared with the generated motion number sequence 33c-0 to determine what action the character is currently taking (step S61). .
[0107]
Thereafter, it is determined whether or not the technique already recognized in 33c-1 is included (step S62). In the case of NO, the motion number sequence of 33c-0 and 21a-7 is matched, and it is determined whether or not the technique has been confirmed (step S63). If it is not confirmed, the process ends. If it is confirmed, 33c is determined. The technique number (the number registered in the knowledge data 21) is recorded in -1 (step S64).
[0108]
On the other hand, if YES in step S62, it is determined whether or not the technique is determined by matching the motion number strings 33c-0j and 21a-7 (step S65). If the technique is not fixed, the timer 33c-3 (interval between techniques) is increased by 1 (step S66), and the process ends.
[0109]
If the technique has been confirmed, it is determined whether the timer 33c-3 (interval between techniques) is smaller than the threshold (step S67). If it is smaller, the actions 33c-1, 33c-2 and 33c-3 are performed. A pattern is recorded as learning data 40 in the SRAM 4 (step S68).
[0110]
The contents of step S68 correspond to step S53 for learning the behavior pattern of FIG. 12, and the details thereof are as shown in FIG. 13 described above.
[0111]
In step S67, when the 33c-3 timer (skill-skill interval) is not smaller than the threshold and when recording is performed as the learning data 40 in step S68, the 33c-3k timer (skill-skill interval) is set. Clear, therefore, 33c-3 = empty string (null) (step S69). Next, 33c-2 is inserted into 33c-1, 33c-2 is cleared, and the process ends.
[0112]
【The invention's effect】
As described above according to the embodiment, the reaction on the computer side changes every time the player plays a game according to the present invention. As a result, with the improvement of the player's progress, the computer side uses more advanced and effective tactics and can keep the player's interest.
[0113]
In addition, since it is possible to adjust the difficulty according to the player by judging the characteristics and skills of the player and influencing the behavior, it is easy for even beginners to tackle. Furthermore, the said Example is an Example of this invention, and this invention is not limited to these Examples. Further, the present invention can be applied to all types of games in which a computer and a player confront each other. Furthermore, when creating a new game as in the past, it is possible to greatly reduce the time and effort of creating a behavior pattern on the computer side in consideration of all situations that may occur during the game.
[Brief description of the drawings]
FIG. 1 is a block diagram of an apparatus according to an embodiment of the present invention.
FIG. 2 is an example of a monitor display screen.
FIG. 3 is a diagram showing an overall operation flow of the game apparatus of the present invention.
FIG. 4 is a diagram showing a detailed flow of a character action determination procedure SS1 of FIG.
FIG. 5 is a diagram illustrating knowledge data.
FIG. 6 is a diagram illustrating behavior pattern data.
FIG. 7 is a diagram illustrating character state transition data.
FIG. 8 is a diagram illustrating learning data.
9 is a diagram illustrating assembly of learning data in FIG. 8. FIG.
FIG. 10 is a flowchart illustrating details (part 1) of step S24 of FIG. 4;
FIG. 11 is a flowchart illustrating details (No. 2) of step S24 in FIG. 4;
FIG. 12 is a diagram showing a detailed flow of the learning procedure SS2 of FIG. 3;
FIG. 13 is a diagram illustrating a learning flow of behavior patterns.
FIG. 14 is a diagram showing an example flow of a learning procedure using a behavior pattern as a cooperative technique.
[Explanation of symbols]
1 CPU body
2, 5 ROM
3 RAM
4 SRAM
6 Input device
7 Output control circuit
8 Output device
10 Character action decision procedure
11 Operation result judgment procedure
12 Learning procedures
13 Output control procedure
20 Game program body
21 Knowledge data
30 Behavioral Paren Data
31 Character state data
32 Player personal data
33 Character state transition data
40 learning data
41 Output data

Claims (4)

An input device for a player to input an operation signal ;
Determining the motion number as the motion information of the first character based on the operation signal from the input device,
Selecting a motion number as movement information of the second character that plays against the first character according to the player's behavior tendency ;
A CPU for controlling a battle game between the first character and the second character;
A display device including a display unit for displaying a motion to be reproduced as an action of the first character and the second character;
Knowledge data associated with a technique that can be taken by the first character and the second character, and a motion number sequence to be reproduced when the technique is taken ;
Each time the state of the first character and the second character changes, the data indicates the state of the character whose value changes correspondingly , and includes at least the coordinate position of the character in the virtual space and the motion number being reproduced. Character state data including data, and
Means the motion number sequence of the first character and the second character, character state transition data for recording the motion number as the motion information of the character read from the character state data in time series,
Player personal data indicating a tendency of the player's behavior stored in a database from the behavior result of the player who is playing a game;
Action pattern data which is data indicating the current action pattern of the second character;
Learning data representing an action pattern defining the technique of the second character and the interval between the techniques;
A memory for storing
The CPU is
Based on the determined actions of the first character and the second character, the state of each character is changed,
Updating the character state data based on the result of the temporary or final victory or defeat determined for the change in the state;
Referring to the character state data, the tendency of the player's behavior is examined, and the player's personal data is updated,
A motion number is read out from the character state data, and recorded in the memory in time series as character state transition data representing a motion number string,
The motion number sequence recorded in the memory as the character state transition data is referred to, and compared with the motion number sequence defined in the knowledge data , the technique of the second character and the interval between the techniques are determined. Record the determined technique and the action pattern defining the technique-to-skill interval in the memory as learning data;
Based on the character state data and the player's personal data, a flag indicating the direction of the skill to be selected as the action of the second character is stored in action pattern data, and based on the flag, Select the behavior pattern of the second character from the knowledge data and learning data ,
Storing the selected action pattern in the memory as action pattern data of the second character;
A computer game device having a learning function. In claim 1,
The learning data to be generated is not an already learned behavior pattern, and if there is no free space in the memory, a pattern having a low priority is erased and recorded in the memory. Computer game device. Game player motion number as the motion information of the first character is determined based on the operation signal from the input device for inputting an operation signal,
Selecting a motion number as movement information of the second character that plays against the first character according to the player 's behavior tendency ;
A CPU for displaying a determined action of the first character and the second character on a display unit to execute a battle game ; a technique that the first character and the second character can take; In a method for determining an action pattern of a character in a computer game device comprising a memory for storing knowledge data associated with a motion number sequence to be reproduced when taking
The CPU is
Each time the state of the first character and the second character is changed, a data indicating the state of the corresponding to the value changes character, and the coordinate position in the virtual space of at least characters, the motion number being played Character state data, including
Means the motion number sequence of the first character and the second character, character state transition data for recording the motion number as the motion information of the character read from the character state data in time series,
Player personal data indicating a tendency of the player's behavior stored in a database from the behavior result of the player who is playing a game;
Action pattern data which is data indicating the current action pattern of the second character;
Learning data representing a technique of the second character and an action pattern defining an interval between the techniques.
Storing in the memory,
Furthermore, the CPU
Based on the prior SL determined action of the first character and the second character, changing the state of each character,
Updating the character state data based on the result of the temporary or final victory or defeat determined for the change in the state;
Referring to the character state data, the tendency of the player's behavior is examined, and the player's personal data is updated,
A motion number is read out from the character state data, and recorded in the memory in time series as character state transition data representing a motion number string,
The motion number sequence recorded in the memory as the character state transition data is referred to, and compared with the motion number sequence defined in the knowledge data , the technique of the second character and the interval between the techniques And the action pattern defining the determined technique and the interval between techniques is recorded as learning data in the memory ,
Based on the character state data and the player personal data, a flag indicating the direction of selecting the skill of the tendency as the action of the second character is stored in the action pattern data, and based on the flag, A method for determining an action pattern of a character in a computer game device, wherein the action pattern of the second character is selected from knowledge data and learning data. In claim 3 ,
In the computer game apparatus, the generated learning data is not a behavior pattern that has already been learned, and if there is no free space in the memory, a low-priority pattern is deleted and recorded in the memory. A method for determining a character's action pattern.

Images