-40-
绗洓绔?/p>
鍔ㄦ�佽鍒?/p>
搂
1
寮曡█
1.1
鍔ㄦ�佽鍒掔殑鍙戝睍鍙婄爺绌跺唴瀹?/p>
鍔ㄦ�佽鍒?/p>
锛?/p>
dynamic programming
锛?/p>
鏄繍绛瑰鐨勪竴涓垎鏀紝
鏄眰瑙ｅ喅绛栬繃绋?/p>
锛?/p>
decision
process
锛?/p>
鏈�浼樺寲鐨勬暟瀛︽柟娉曘�?/p>
20
涓栫邯
50
骞翠唬鍒?/p>
R. E. Bellman
绛変汉鍦ㄧ爺绌跺闃舵鍐崇瓥杩?/p>
绋?/p>
(multistep
decision
process)
鐨勪紭鍖栭棶棰樻椂锛屾彁鍑轰簡钁楀悕鐨勬渶浼樻�у師鐞嗭紙
principle
of
optimality
锛?/p>
锛屾妸澶氶樁娈佃繃绋嬭浆鍖栦负涓�绯诲垪鍗曢樁娈甸棶棰橈紝閫愪釜姹傝В锛屽垱绔嬩簡瑙ｅ喅杩欑被杩囩▼
浼樺寲闂鐨勬柊鏂规硶鈥斿姩鎬佽鍒掋�?/p>
1957
骞村嚭鐗堜簡浠栫殑鍚嶈憲銆?/p>
Dynamic
Programming
銆?/p>
锛岃繖
鏄棰嗗煙鐨勭涓�鏈憲浣溿�?/p>
鍔ㄦ�佽鍒掗棶涓栦互鏉ワ紝
鍦ㄧ粡娴庣鐞嗐�?/p>
鐢熶骇璋冨害銆?/p>
宸ョ▼鎶�鏈拰鏈�浼樻帶鍒剁瓑鏂归潰寰楀埌浜嗗箍
娉涚殑搴旂敤銆備緥濡傛渶鐭矾绾裤�佸簱瀛樼鐞嗐�佽祫婧愬垎閰嶃�佽澶囨洿鏂般�佹帓搴忋�佽杞界瓑闂锛岀敤鍔?/p>
鎬佽鍒掓柟娉曟瘮鐢ㄥ叾瀹冩柟娉曟眰瑙ｆ洿涓烘柟渚裤�?/p>
铏界劧鍔ㄦ�佽鍒掍富瑕佺敤浜庢眰瑙ｄ互鏃堕棿鍒掑垎闃舵鐨勫姩鎬佽繃绋嬬殑浼樺寲闂锛?/p>
浣嗘槸涓�浜涗笌鏃?/p>
闂存棤鍏崇殑闈欐�佽鍒掞紙濡傜嚎鎬ц鍒掋�侀潪绾挎�ц鍒掞級
锛屽彧瑕佷汉涓哄湴寮曡繘鏃堕棿鍥犵礌锛屾妸瀹冭涓?/p>
澶氶樁娈靛喅绛栬繃绋嬶紝涔熷彲浠ョ敤鍔ㄦ�佽鍒掓柟娉曟柟渚垮湴姹傝В銆?/p>
搴旀寚鍑猴紝鍔ㄦ�佽鍒掓槸姹傝В鏌愮被闂鐨勪竴绉嶆柟娉曪紝鏄�冨療闂鐨勪竴绉嶉�斿緞锛岃�屼笉鏄?/p>
涓�绉嶇壒娈婄畻娉曪紙濡傜嚎鎬ц鍒掓槸涓�绉嶇畻娉曪級
銆傚洜鑰岋紝瀹冧笉璞＄嚎鎬ц鍒掗偅鏍锋湁涓�涓爣鍑嗙殑鏁?/p>
瀛﹁〃杈惧紡鍜屾槑纭畾涔夌殑涓�缁勮鍒欙紝
鑰屽繀椤诲鍏蜂綋闂杩涜鍏蜂綋鍒嗘瀽澶勭悊銆?/p>
鍥犳锛?/p>
鍦ㄥ涔?/p>
鏃讹紝
闄や簡瑕佸鍩烘湰姒傚康鍜屾柟娉曟纭悊瑙ｅ锛?/p>
搴斾互涓板瘜鐨勬兂璞″姏鍘诲缓绔嬫ā鍨嬶紝
鐢ㄥ垱閫犳�х殑
鎶�宸у幓姹傝В銆?/p>
渚?/p>
1
鏈�鐭矾绾块棶棰?/p>
涓嬮潰鏄竴涓嚎璺綉锛?/p>
杩炵嚎涓婄殑鏁板瓧琛ㄧず涓ょ偣涔嬮棿鐨勮窛绂?/p>
锛堟垨璐圭敤锛?/p>
銆?/p>
璇曞姹備竴鏉＄敱
A
鍒?/p>
G
璺濈鏈�鐭紙鎴栬垂鐢ㄦ渶鐪侊級鐨勮矾绾裤�?/p>
渚?/p>
2
鐢熶骇璁″垝闂
宸ュ巶鐢熶骇鏌愮浜у搧锛屾瘡鍗曚綅锛堝崈浠讹級鐨勬垚鏈负
1
锛堝崈鍏冿級
锛屾瘡娆″紑宸ョ殑鍥哄畾鎴愭湰涓?/p>
3
锛堝崈鍏冿級
锛屽伐鍘傛瘡瀛ｅ害鐨勬渶澶х敓浜ц兘鍔涗负
6
锛堝崈浠讹級
銆傜粡璋冩煡锛屽競鍦哄璇ヤ骇鍝佺殑闇�姹傞噺绗?/p>
涓�銆佷簩銆佷笁銆佸洓瀛ｅ害鍒嗗埆涓?/p>
2
锛?/p>
3
锛?/p>
2
锛?/p>
4
锛堝崈浠讹級
銆傚鏋滃伐鍘傚湪绗竴銆佷簩瀛ｅ害灏嗗叏骞寸殑闇�
姹傞兘鐢熶骇鍑烘潵锛岃嚜鐒跺彲浠ラ檷浣庢垚鏈紙灏戜粯鍥哄畾鎴愭湰璐癸級
锛屼絾鏄浜庣涓夈�佸洓瀛ｅ害鎵嶈兘涓?/p>
甯傜殑浜у搧闇�浠樺瓨鍌ㄨ垂锛?/p>
姣忓姣忓崈浠剁殑瀛樺偍璐逛负
0.5
锛堝崈鍏冿級
銆?/p>
杩樿瀹氬勾鍒濆拰骞存湯杩欑浜у搧
鍧囨棤搴撳瓨銆傝瘯鍒跺畾涓�涓敓浜ц鍒掞紝
鍗冲畨鎺掓瘡涓搴︾殑浜ч噺锛?/p>
浣夸竴骞寸殑鎬昏垂鐢?/p>
锛堢敓浜ф垚鏈?/p>
鍜屽瓨鍌ㄨ垂锛夋渶灏戙�?/p>
1.2
鍐崇瓥杩囩▼鐨勫垎绫?/p>
鏍规嵁杩囩▼鐨勬椂闂村彉閲忔槸绂绘暎鐨勮繕鏄繛缁殑锛屽垎涓虹鏁ｆ椂闂村喅绛栬繃绋嬶紙
discrete-time
decision process
锛夊拰杩炵画鏃堕棿鍐崇瓥杩囩▼锛?/p>
continuous-time decision process
锛?/p>
锛涙牴鎹繃绋嬬殑
婕斿彉鏄‘瀹氱殑杩樻槸闅忔満鐨勶紝鍒嗕负纭畾鎬у喅绛栬繃绋嬶紙
deterministic
decision
process
锛夊拰闅?