| Project/Area Number |
23500288
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Research Field |
Sensitivity informatics/Soft computing
|
|---|
| Research Institution | Tokyo Denki University |
|---|
Principal Investigator |
SETSUO Tsuruta 東京電機大学, 情報環境学部, 教授 (00366395)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
SAKURAI Yoshitaka 明治大学, 総合数理学部, 准教授 (30408653)
|
|---|
| Co-Investigator(Renkei-kenkyūsha) |
TERANO Takao 東京工業大学, 総合理工学研究科(研究院), 教授 (20227523)
KITA Hajime 東京工業大学, 学術情報メディアセンター, 教授 (20195241)
IKEDA Kokoro 北陸先端科学技術大学院大学, 情報科学研究科, 准教授 (80362416)
|
|---|
| Project Period (FY) |
2011-04-28 – 2015-03-31
|
| Project Status |
Completed (Fiscal Year 2014)
|
| Budget Amount *help |
¥5,070,000 (Direct Cost: ¥3,900,000、Indirect Cost: ¥1,170,000)
Fiscal Year 2014: ¥1,040,000 (Direct Cost: ¥800,000、Indirect Cost: ¥240,000)
Fiscal Year 2013: ¥1,040,000 (Direct Cost: ¥800,000、Indirect Cost: ¥240,000)
Fiscal Year 2012: ¥1,040,000 (Direct Cost: ¥800,000、Indirect Cost: ¥240,000)
Fiscal Year 2011: ¥1,950,000 (Direct Cost: ¥1,500,000、Indirect Cost: ¥450,000)
|
|---|
| Keywords | 遺伝的アルゴリズム / 強化学習 / 自律協調 / 事例ベース / 人間中心 / GA / 文化遺伝子 / GA：遺伝的アルゴリズム / 配送 / 最適化 / 事例 / 人間的制約 / 多様性 / 進化知能 / 自律分散 / 進化型知能化方式 / 問題パターン / 分散並列進化方式 / 多様性エントロピー / パラメータ制御 / 適応的制御 / ヒューリスティクス / メタヒューリスティクス |
|---|
| Outline of Final Research Achievements |
To control huge amount of parameters in Genetic Algorithm (GA) for delivery route optimization, GA introducing the reinforcement learning was proposed. This maximizes the long-range rewards using GA’s characteristic of multiple points for search. However, the performance was not sufficient. Thus cases with local but human orient adjustment heuristics NI (Nearest Insertion) were introduced into GA, due to the insight that real problems are similar to former problems. Solutions can be derived from similar former solutions, considering human oriented factors (e.g. inheriting most of delivery routes). Experimental evaluation revealed remarkable results. Even though the fastest effective TSP solving method LKH needed more than 3 seconds, the proposed method yielded results within 3% of the worst error rate and in less than 3 seconds. Furthermore, the proposed method can inherit many delivery route orders, while LKH tends to make reverse order of routes in delivery route optimization.
|
