プロダクションシステム

  ワーキングメモリ(WM):     項の配列(集合)
  ルールベース(RB):         ルールの連想配列(一般オブジェクト)

    但し
      項:           [ 述語, 対象, … ] （配列）
        述語・対象: 文字列
      ルール:       { condition: 条件, action: 動作 }
        条件:       [ 項, … ] （項の配列(集合)）
        動作:       [ 項, … ] （項の配列(集合)）

  (cf. term_set.js)

例1: 積木の世界 (前向き推論の例)

  (cf. forward_reasoning.js)

  WM要素
    ["on-table", X] : 積木Xがテーブルの上に(接して)のっている
    ["on", X, Y]    : 積木Xが積木Yの上に積まれている
    ["clear", X]    : 積木Xの上には何も積まれていない

  初期状態

      C
      A B
    -------
    [["on-table", "a"], ["on", "c", "a"], ["clear", "c"], ["on-table", "b"], ["clear", "b"]]

  目標状態

      A
      B
      C
    -----
    [["on-table", "c"], ["on", "b", "c"], ["on", "a", "b"], ["clear", "a"]]

  ルールベース
    sXY --- X を Y に積む
      sab: { condition: [["on-table", "a"], ["clear", "a"], ["clear", "b"]],
             action:    [["clear", "a"], ["on", "a", "b"]]},
      sac: { condition: [["on-table", "a"], ["clear", "a"], ["clear", "c"]],
             action:    [["clear", "a"], ["on", "a", "c"]]},
      sbc: { condition: [["on-table", "b"], ["clear", "b"], ["clear", "c"]],
             action:    [["clear", "b"], ["on", "b", "c"]]},
      sba: { condition: [["on-table", "b"], ["clear", "b"], ["clear", "a"]],
             action:    [["clear", "b"], ["on", "b", "a"]]},
      sca: { condition: [["on-table", "c"], ["clear", "c"], ["clear", "a"]],
             action:    [["clear", "c"], ["on", "c", "a"]]},
      scb: { condition: [["on-table", "c"], ["clear", "c"], ["clear", "b"]],
             action:    [["clear", "c"], ["on", "c", "b"]]},

    dXY --- X を Y から下ろす
      dab: { condition: [["clear", "a"], ["on", "a", "b"]],
             action:    [["on-table", "a"], ["clear", "a"], ["clear", "b"]]},
      dac: { condition: [["clear", "a"], ["on", "a", "c"]],
             action:    [["on-table", "a"], ["clear", "a"], ["clear", "c"]]},
      dbc: { condition: [["clear", "b"], ["on", "b", "c"]],
             action:    [["on-table", "b"], ["clear", "b"], ["clear", "c"]]},
      dba: { condition: [["clear", "b"], ["on", "b", "a"]],
             action:    [["on-table", "b"], ["clear", "b"], ["clear", "a"]]},
      dca: { condition: [["clear", "c"], ["on", "c", "a"]],
             action:    [["on-table", "c"], ["clear", "c"], ["clear", "a"]]},
      dcb: { condition: [["clear", "c"], ["on", "c", "b"]],
             action:    [["on-table", "c"], ["clear", "c"], ["clear", "b"]]}

例2: 旅行案内 (後向き推論の例)

  概要
    後向き推論の木を深さ優先探索する (バックトラックあり)
      節点:   ワーキングメモリのプロパティ backtrace に適用ルール(の名前)のリスト(配列)を与えたもの
      根節点: 目標状態の WM に，プロパティ backtrace として [] を与えたもの
      子節点: 親節点のワーキングメモリにルールを逆向きに適用して得られる WM
              プロパティ backtrace のリスト(配列)には，使用されたルール(の名前)が(前から)追加される

  (cf. backward_reasoning.js)

  ルールベース
    旅行先決定ルール
    r1:  { condition: [["usa"], ["english"]],       action: [["honolulu"]]  },
    r2:  { condition: [["europe"], ["france"]],     action: [["paris"]]     },
    r3:  { condition: [["usa"], ["continent"]],     action: [["losange"]]   },
    r4:  { condition: [["island"], ["equator"]],    action: [["honolulu"]]  },
    r5:  { condition: [["asia"], ["equator"]],      action: [["singapore"]] },
    r6:  { condition: [["island"], ["micronesia"]], action: [["guam"]]      },
    r7:  { condition: [["swimming"]],               action: [["equator"]]   },

    旅行先決定時ルール
    rg1: { condition: [["honolulu"]],  action: [["honolulu"], ["ok"]]  },
    rg2: { condition: [["paris"]],     action: [["paris"], ["ok"]]     },
    rg3: { condition: [["losange"]],   action: [["losange"], ["ok"]]   },
    rg4: { condition: [["singapore"]], action: [["singapore"], ["ok"]] },
    rg5: { condition: [["guam"]],      action: [["guam"], ["ok"]]      }

  目標状態
    [["ok"]]