Received May 16, 2013/ revised July 8, 2013/ accepted August 22, 2013
Copyright ⵑ 2013 by the Korean Society of Civil Engineers
ǣǣȀȀǤǤȀͳͲǤͳʹͷʹȀ ǤʹͲͳ͵Ǥ͵͵ǤǤʹͶͷ Ǥ ǤǤ
DSRC ⵎẊἺ#ⴲⱧ㬚#ኞ❋ᦂḚ#ᢦᏮ#㝳㭇⢚ᇂ#⯆㏟
෴ச ȵୋ׆
Kim, Hyungjoo*, Jang, Kitae**
Short-Term Prediction of Travel Time Using DSRC on Highway
ABSTRACT
This paper develops a travel time prediction algorithm that can be used for real-time application. The algorithm searches for the most similar pattern in historical travel time database as soon as a series of real-time data become available. Artificial neural network approach is then taken to forecast travel time in the near future. To examine the performance of this algorithm, travel time data from Gyungbu Highway were obtained and the algorithm is applied. The evaluation shows that the algorithm could predict travel time within 4% error range if comparable patterns are available in the historical travel time database. This paper documents the detailed algorithm and validation procedure, thereby furnishing a key to generating future travel time information.
Key words : Real-time traffic information, Travel time prediction, DSRC, Artificial neural network, Random number
Ⅹಾ
⩥ᰍʭḡ☖⧪eᩩ⊂ŝšಉࡽ݅᧲⦽ᩑǍॅᯕᙹ⧪ࡹᨩḡอ, ⦽ǎŁᗮࠥಽ✚ᖒᨱ฿۵ᩩ⊂ႊჶುᨱݡ⦽ᝅ᷾ᩑǍ۵ᇡ᳒⦽ᝅᱶᯕ݅.
ᯕᨱᅙᩑǍᨱᕽ۵ᝅᱽ☖⧪eᮥʑၹᮝಽ⦽DSRC ᯱഭෝၵ┶ᮝಽ⦽ǎŁᗮࠥಽᨱᱢᱩ⦽ᩩ⊂ႊჶುᮥࠥ⇽⦽݅. ĞᇡŁᗮࠥಽᦩᖒ JC~᪅ᔑIC Ǎe᮹24eDSRC ᯱഭෝᯕᬊ⦹໑݉ᵝʑ☖⧪eᩩ⊂ၰእᖁ⩶šĥᨱᕽ׳ᮡᱶ⪶ࠥෝᅕᯕ۵ᯙŖᝁĞʑჶᮥᱢᬊ
⦽݅. ᯕᨕᕽ߅ӽᙹෝᯕᬊ⦽☖⧪eᩩ⊂đŝ᮹ᱶ⪶ࠥá᷾ᮥᝅ⦽݅. ☖⧪eᩩ⊂đŝ᪅₉ᮉᯕ᧞4%ಽᬑᙹ⦽ᩩ⊂ಆᮥᅕᩡᮝ ໑, ᯕ۵➉▕ʑၹᯙŖᝁĞᩩ⊂ᯕಆᯱഭ᮹ᱥญŝᱶŝ↽ᱢ᮹᯦ಆ⊖ၰᮡܪ⊖᮹ᖁᱶᮝಽᯙ⦽đŝಽ❱݉ࡽ݅. ☖⧪eᩩ⊂đ ŝ᮹á᷾ᮥ᭥⧕ᕽ߅ӽᙹෝᯕᬊ⦹ᩡᮝ໑, ߅ӽᙹaᯕಆᯱഭ➉▕ᨱ⡍⧉ࡹḡᦫᮡĞᬑᝅ⊂⊹᪡᮹᪅₉ᮉᯕ18.98%ಽ׳íࠥ⇽ࡹᨩ
݅. ᯕ۵ᯙŖᝁĞᮥᯕᬊ⦽☖⧪eᩩ⊂➉▕DBaᩩ⊂᮹ᱶ⪶ࠥᨱᵝ⦹í᯲ᬊ⦽đŝಽ❱݉ࡽ݅. ᅙᩑǍ᮹đŝෝ☖⧕ᕽ⦽ǎŁ ᗮࠥಽ✚ᖒᨱ฿۵☖⧪eᩩ⊂ၰᱶᅕᱽŖᯕa܆⧁äᮝಽ❱݉ࡽ݅.
áᔪᨕ ᝅeƱ☖ᱶᅕ, ☖⧪eᩩ⊂, ݉Ñญᱥᬊ☖ᝁ, ᯙŖᝁĞ, ߅ӽᙹ
1. ᕽು
ḡ܆⩶ Ʊ☖ℕĥ(Intelligent Transportation System, ᯕ⦹ ITS)᮹ ᵲ⦽ ⦹᭥ᜅ▽ᯙ ℉݉ᩍ⧪ᯱƱ☖ᱶᅕᜅ▽(Advanced Traveller Information Systems, ᯕ⦹ ATIS)ᮡ ᬑญӹᨱ ITS ʑᅙĥ⫮ᯕ ᙹพࡽ ᯕ⬥ ḡɩʭḡ ⪽ၽ⦽ ᩑǍa ḥ⧪ࡹŁ ᯩ݅.
ATIS ᇥᨱᯩᨕ☖⧪eᱶᅕ۵ᙹḲࡽƱ☖ᯱഭෝᯕᬊ⦹ᩍ☖⧪(travel)ŝšಉࡽƱ☖᳑Õᮥᩩ⊂⦹ᩍᩍ⧪ᯱᨱíᱶᅕᱽŖᮥ
༊ᱢᮝಽ ⦽݅. ᝅe Ʊ☖ᱶᅕ ᱽŖᮡ ᬕᱥᯱॅᨱí ↽ᱢ᮹ Ğಽᖁ┾ᮥ ᮁࠥ⦹ᩍ ᯕᬊᯱ ⠙᮹᪡ Ʊ☖ᦩᱥᮥ ᱽŖ⦹ʑ ᭥⧉ᯕ݅.
⩥ᰍŁᗮࠥಽᔢ᮹Ʊ☖ᱶᅕ۵ݡᇡᇥ⥥áḡʑၰ݉Ñญᱥᬊ☖ᝁ(Dedicated Short Range Communications, ᯕ⦹DSRC)ᮥ
İࣀėॡ
ᯕᬊ⦹ᩍᙹḲࡹŁᯩ݅. ⥥áḡʑ᮹Ğᬑ₉పáḡᨱᯩᨕ׳ᮡ
ᱶ⪶ࠥෝaḡ۵ᰆᱱᯕᯩḡอ, áḡʑ᪅᯲࠺ᮝಽᯙ⦽ᅖǍ
᪅eᯕᗭࡹ۵݉ᱱᯕᯩ݅. ੱ⦽☖⧪eᔑᱶ⦹ඹᇡ᮹
ᱶℕ ၰ ᔍŁಽ ᯙ⦽ ᩢ⨆ᯕ ᪽łࢁ ᙹ ᯩᨕ ☖⧪e ᱶᅕ᮹
ᝁᖒᮥᅕᰆ⧁ᙹᨧ݅. ၹ໕, DSRC ᯱഭෝᯕᬊ⦽Ǎeáḡ᮹
Ğᬑᝅᱽ☖⧪eᮥᔑᱶ⦹ᩍᝁᖒ׳ᮡƱ☖ᱶᅕෝᱽŖ⦽݅.
⩥ᰍʭḡ☖⧪eᩩ⊂ŝšಉࡽ݅᧲⦽ᩑǍॅᯕᙹ⧪ࡹᨩḡ อ, ⦽ǎŁᗮࠥಽ✚ᖒᨱ฿۵ᩩ⊂ႊჶುᨱݡ⦽ᝅ᷾ᩑǍ۵ᇡ᳒
⦽ᝅᱶᯕ݅. ੱ⦽ᩩ⊂☖⧪eᨱݡ⦽ᱶ⪶ࠥá᷾ᯕᙹ⧪ࡹḡ
༜⧕ ᇥᕾđŝ᮹ ᝁᖒᮥ ᅕᰆ⦹ḡ ༜⦹۵ ⦽ĥa ᳕ᰍ⦽݅.
ᯕᨱᅙᩑǍᨱᕽ۵ᝅᱽ☖⧪eᮥʑၹᮝಽ⦽DSRC ᯱഭෝ
ၵ┶ᮝಽ ⦽ǎŁᗮࠥಽᨱ ᱢᱩ⦽ ᩩ⊂ႊჶುᮥ ࠥ⇽⦽݅. ᯕෝ
᭥⦹ᩍ ⦽ǎࠥಽŖᔍᨱᕽᙹḲࡽ ĞᇡŁᗮࠥಽᦩᖒJC~᪅ᔑIC Ǎe᮹ 24e DSRC ᯱഭෝ ᯕᬊ⦹໑, ݉ᵝʑ☖⧪e ᩩ⊂
ၰእᖁ⩶(non-linear) šĥᨱᕽ׳ᮡᱶ⪶ࠥෝᅕᯕ۵ᯙŖᝁĞ
(Artificial Neural Network)ᮥᱢᬊ⦽݅. ᯕᨕᕽᩩ⊂☖⧪e᮹
ᱶ⪶ࠥá᷾ᮥ᭥⦹ᩍ߅ӽᙹ(random number)ෝᯕᬊ⦹ᩍእ Ʊᇥᕾᮥᝅ⦽݅. ษḡสᮝಽđುᨱᕽ۵ᅙᩑǍ᮹ᖒŝၰ
⦽ĥ, əญŁ ⨆⬥ ᩑǍႊ⨆ᮥ ࠥ⇽⦽݅.
2. šಉᯕುၰᖁ⧪ᩑǍŁₑ
2.1 վլ࠷
↽Ⅹ᮹ᝁĞ⫭ಽ}ֱᮡ1943֥ၙǎ᮹Warren McCulloch
᪡ᙹ⦺ᯱWalter Pittsᨱ᮹⧕Łᦩࡹᨩ݅. ᝁĞᮡ⍕⥉░a
ᔍ௭᮹⦺܆ಆᮥᙹ⧪⦹Łᯱ}ၽࡹᨩᮝ໑, ᙹ⦺ᱢŁญ᷹ᮝ ಽǍᖒࡹᨕᯩ݅. ᝁĞᮡەᨱᗮ⦹۵ᮁܼ(unit)ॅŝᮁܼ
ᔍᯕ᮹ᩑđvࠥ(weight)ಽǍᖒࡹᨕᯩᮝ໑, ĥ⊖᮹ᙹᨱ
݉⊂ᝁĞŝ݅⊖ᝁĞᮝಽǍᇥࡽ݅. ᝁĞ༉⩶ᨱᕽە᮹
ᵝʑ܆ᮡ᯦ಆŝᩑđvࠥ᮹aᵲ⧊(NET)ᮥǍ⦽⬥⪽ᖒ⪵
⧉ᙹᨱ ᮹⧕ ⇽ಆᮥ ԕᅕԕ۵ äᯕ݅. ⪽ᖒ⪵ ⧉ᙹ(Activation function)ᮡ݉᳑᷾a⦹۵⧉ᙹᯕᨕ⦹໑, ᯝၹᱢᮝಽ݉ɚᖒ/᧲
ɚᖒ ⧉ᙹ, ᖁ⩶/እᖁ⩶ ⧉ᙹ, ᩑᗮ/ᯕḥ ⧉ᙹ ॒ᮝಽ ᇥඹ⦽݅.
ᝁĞ༉⩶ᨱᕽᩎᱥ❭⦺ᯕ, ✚ᱶ⦽᮲ᬊ༊ᱢᨱᱢ⧊⦹ࠥ
ಾەe᮹ᩑđvࠥෝᱢ᮲┅۵ŝᱶᮝಽ⩥ᰍᝁĞ༉⩶ᨱ ᕽaᰆձญᯕᬊࡹ۵ߙ┡⦺(Delta learning)ᮡ݅⊖ᝁĞ᮹
⦺Ƚ⊺ᮝಽᵝಽᔍᬊࡽ݅. ᯕ౨í݅⊖ᝁĞŝߙ┡Ƚ⊺ᮥ
đ⧊⦽⩶┽ෝᩎᱥ❭Łญ᷹(Backpropagation algorithm)ᯕ
ᇡ໑, ᩑᗮ⪽ᖒ⪵⧉ᙹอᮥᔍᬊ⦽݅. ߙ┡⦺ᮡ⦺ᝁ⪙ಽᕽ
༊⢽⊹᪡ᝅᱽ⇽ಆ⊹᮹₉ᯕsŝ⪽ᖒ⪵⧉ᙹ᮹ၙᇥsᯕᔍᬊࡹ
۵ ᱱᯕ ✚Ḷᯕ݅.
2.2 ധෘਏԩુ౸ࡦ෴ճఝ
☖⧪eᩩ⊂༉⩶ᮡⓍí☖ĥᱢႊჶŝဍ౩ᯕᖹႊჶᮝಽ
ӹ٭ᙹᯩ݅. ☖ĥᱢႊჶᮡᯕಆᯱഭၰᝅeᯱഭෝᯕᬊ⦹ᩍ
ᩩ⊂☖⧪eᮥࠥ⇽⦽݅. ၹ໕ᨱဍ౩ᯕᖹ༉⩶ᮡƱ☖ᔢ⫊
ၰᱶℕಽᯙ⦽∊Ċ❭ᗮࠥᨱʑၹ⦹ᩍ☖⧪eᩩ⊂ᮥᙹ⧪⦽݅.
☖ĥᱢႊჶ᮹Ğᬑ⫭ȡᇥᕾ༉⩶, ĥᩕ༉⩶, ᯙŖᝁĞ, ⋝อ
⦥░॒ᯕᔍᬊࡽ݅. ⫭ȡᇥᕾ༉⩶ᮡᯝၹᱢ(Stationary) ᔢ⫊᮹
݉ʑᩩ⊂ᨱ۵ ᬑᙹ⦽ ᩩ⊂ಆᮥ ᅕᯕḡอ, ᮁŁ ၰ ᱶℕಽ ᯙ⦽
Ʊ☖⩥ᔢᨱᕽ۵ԏᮡᱶ⪶ࠥෝaḥ݅. ĥᩕ༉⩶ᮡᗮࠥ, Ʊ☖ప, ᱱᮁᮉ॒᮹ĥᩕᱶᅕෝᔍᬊ⦹۵äᮝಽARIMA ༉⩶ᯕݡ⢽
ᱢᮝಽᔍᬊࡹŁᯩᮝ໑, ĥᱩᱢᄡ࠺ᨱᬑᙹ⦽ᩩ⊂ಆᮥaḥ݅.
⦹ḡอᔍŁၰᱶℕಽᯙ⦽Ʊ☖⩥ᔢᨱᕽ۵ԏᮡᝁᖒᮥaḡ۵
݉ᱱᯕᯩ݅. ᯙŖᝁĞ༉⩶ᮡእᖁ⩶☖ĥᱢʑჶᮝಽ⨆ᔢࡽ
ḡࠥ⦺ ႊჶᮥ ☖⧕ ↽ᱢ᮹ đŝෝ ࠥ⇽⦽݅. ੱ⦽ ☖⧪e
ᩩ⊂ᔍŁၰᱶℕ⩥ᔢᨱᕽࠥᬑᙹ⦽ᩩ⊂ಆᮥaḥ݅. ⋝อ⦥░
۵݉ᯝeݡᬑᙹ⦽ᩩ⊂ಆᮥᅕᯕŁᯩᮝ໑, ᙹ⧪ᗮࠥaෙ
ᰆᱱᯕᯩ݅. ⦹ḡอᯙŖᝁĞᨱእ⧕ᩩ⊂ಆᯕԏᮝ໑, እᖁ⩶
ᜅ▽ᯙ Ğᬑ ᩩ⊂ᨱ ᨕಅᬡᯕ ᳕ᰍ⦽݅.
ဍ౩ᯕᖹ༉⩶ᮡÑᱢ༉⩶(macroscopic models), ᵲeᱢ
༉⩶(mesoscopic models), ၙᱢ(microscopic models) ॒ᮝಽ
ӹ٭ᙹᯩ݅. Ñᱢ༉⩶ᮡƱ☖ప, ၡࠥ, əญŁ⠪Ɂᗮ॒ࠥ᮹
Ʊ☖ඹ✚ᖒᨱʑၹ⦹ᩍᇥᕾᮥᝅ⦹۵äᮝಽLWR ༉⩶ၰ
METANET ༉⩶ᯕ ݡ⢽ᱢᮝಽ ᔍᬊࡹŁ ᯩ݅. ᵲeᱢ ༉⩶ᮡ
Ʊ☖༉ᙹ᮹⠪Ɂsŝ}ᄥ₉ప⧪┽ෝ⧉̹Łಅ⦹۵äᮝಽÑ
ᱢ༉⩶ŝၙᱢ༉⩶᮹ᵲeᱢᖒĊᮥaḥ݅. ᵲeᱢ༉⩶ᮡ
}ᄥ₉పᮥဍ౩ᯕᖹ⧁ᙹᯩḡอ, ₉ప᮹ᬡḢᯥၰၹ᮲ᮡ
ยⓍ݉᭥᮹ Ḳĥsᯕ ᱢᬊࡽ݅. ݡ⢽ᱢᮝಽ MITᨱᕽ }ၽࡽ
DYNAMIT ༉⩶ᯕձญᔍᬊࡹŁᯩ݅. ၙᱢ༉⩶ᮡ}ᄥ₉ప᮹
࠺ᱢᯙᬡḢᯥᮥ༉ࢱ⡍⧉⦹۵ᰆᱱᮥaḥ݅. ₉ప✚ᖒ(₉పʙᯕ,
↽ݡaᗮ܆ಆ॒), ₉పᬡḢᯥ✚ᖒ(ᗮࠥ, e, Ñญ॒), ᬕᱥᯱ
⧪┽(₉ప ⇵᳦, ₉ᖁᄡĞ ॒)ෝ ༉ࢱ ⡍⧉⦹໑, ݡ⢽ᱢᮝಽ
CA(Cellular Automata)༉⩶ᯕ ᔍᬊࡽ݅.
ᯕᨱᅙᩑǍᨱᕽ۵☖ĥᱢႊჶ᮹ᯝ⪹ᯙᯙŖᝁĞᮥᯕᬊ⦽
☖⧪eᩩ⊂ᨱݡ⦽ᩑǍෝᝅ⦽݅. ᯙŖᝁĞ᮹Ğᬑ☖⧪
eᩩ⊂ᮁŁၰᱶℕ⩥ᔢᨱᕽࠥᬑᙹ⦽ᩩ⊂ಆᮥaḡ໑, ݉ᵝʑ
☖⧪eᩩ⊂ᨱ׳ᮡᱶ⪶ࠥෝᅕᯕŁᯩᨕ⦽ǎŁᗮࠥಽ✚ᖒᨱ
ᱢ⧊⦹݅Ł ❱݉ࡽ݅.
2.3 ֝ٛ·টෘ֜ճఝ
ᯙŖᝁĞᮡእᖁ⩶☖ĥᱢ༉⩶ᮝಽᩑᗮඹၰ݉ᗮඹǍe
☖⧪eᩩ⊂ᨱձญᱢᬊࡽ݅. Kang and NamKoong (2002) ᩑǍᨱᕽ۵Łᗮࠥಽ☖⧪ഭᙹԊᜅ▽(Toll Collection System,
ᯕ⦹ TCS)ᨱᕽᙹḲࡹ۵ᯱഭ᮹ᗮᖒŝĥᩕᱢ➉▕ᮥȽ
⦹ᩡᮝ໑, ☖⧪eᩩ⊂ᨱ༉ऩᝁĞ༉⩶(Modular Neural Network Model)ᮥ ᱽᦩ⦹ᩡ݅. ༉ऩᝁĞ༉⩶ᮡ݅⊖⟝ᖪ✙
ುᝁĞᮝಽݡȽ༉᯦ಆᯱഭෝ☖⦽☖⧪eᩩ⊂⦺e ᯕʙᨕḡ۵݉ᱱᮥᅕᦩ⦹ʑ᭥⧕እ⦽᯦ಆᯱഭෝᇥඹ⦹ᩍ
⦺ᮥḥ⧪┅۵Ǎ᳑ᯕ݅. ᇥᕾđŝ݉Ñญ᪡ᰆÑญ༉ࢱᩩ⊂
a܆⦽ eᱢ ჵ᭥ෝ ࠥ⇽⦹ᩡᮝ໑, 1}᮹ ᝁĞᮝಽ ⦺⦽
↽ݡ/↽ᗭ᮹ ᩩ⊂ჵ᭥ԕᨱᕽ ࠺ᯝ⦽ ᙹᵡ᮹ ᩩ⊂ಆᮥ ᅕᩡ݅.
Kim and Kim (2001) ᩑǍᨱᕽ۵݉ᗮඹࠥಽᨱᕽ☖⧪e
ᱶᅕᱽŖᮥ᭥⧕⟝ḡᯕುŝᯙŖᝁĞ᮹⧊ᖒ༉⩶ᯙFALEM (Fuzzy Adaptive Learning Estimator for travel time from Multi- information sources)ᮥᱢᬊ⦹ᩡ݅. ḡᱱၰǍeáḡℕĥ ᨱᕽ ᙹḲࡽ áḡᯱഭෝ ⪽ᬊ⦹ᩡᮝ໑, ┡ ༉⩶ŝ እƱෝ ☖⦽
ᮁᬊᖒá᷾ᮥᝅ⦹ᩡ݅. ⦹ḡอ}ၽ༉⩶᮹↽ᗭ⢽ᅙᯥ᮹ᖒྙ
ᱽ᪡ᔢᯕ⦽☖⧪➉▕᮹ᱱᮁᮉญ॒ᨱݡ⦽⦽ĥᱱᯕ᳕ᰍ⦽݅.
Lee (2009) ᩑǍᨱᕽ۵ⓕ్ᜅ░ยႊჶŝᯙŖᝁĞ༉⩶ᮥ
đ⧊⦹ᩍ☖⧪eᩩ⊂༉⩶ᮥ}ၽ⦹ᩡ݅. ᩩ⊂༉⩶᮹᳦ᗮᄡᙹ ۵ ქᜅ᮹ ᝅe ☖⧪eᯕ໑, VDS, GPS, əญŁ ᔍŁᯱഭ
॒ᯕࠦพᄡᙹಽ⪽ᬊࡹᨩ݅. ᯙŖᝁĞᮥᯕᬊ⦽bยⓍ᮹ᩩ⊂
☖⧪eŝᝅᱽ☖⧪e᮹MAPE ᇥᕾđŝ᧞5%᮹᪅₉ᮉᮥ
ᅕᩡ݅.
Innamaa (2005) ᩑǍᨱᕽ۵ࠥᇡŁᗮࠥಽ᮹☖⧪eᯕಆ
ᯱഭෝၵ┶ᮝಽfeedforward multilayer perceptron (MLP) ᯙŖ ᝁĞᮥᱢᬊ⦹ᩍ☖⧪eᩩ⊂ᮥᝅ⦹ᩡ݅. bยⓍ᮹ᩩ⊂
☖⧪eŝᝅᱽ☖⧪e᮹MAPE ᇥᕾđŝ᧞6%᮹᪅₉ᮉᮥ
ᅕᩡ݅. ⦹ḡอᱽᦩࡽ༉⩶᮹Ğᬑᝅeᯱഭ᪡ᩑĥᯕಆᯱഭ ᮹᪅ඹ⊹ᱽÑᨱᨕಅᬡᯕ᳕ᰍ⦹໑, ᯙŖᝁĞ᮹⦺ŝᱶᨱᕽ
Ʊ☖ ⪝ᰂᔢ⫊ᮥ ၹᩢ⦹ḡ ༜⦹۵ ⦽ĥෝ aḥ݅.
Jeong and Rilett (2004) ᩑǍᨱᕽ۵ Automatic Vehicle Location (AVL) ᯱഭ ၰ ᯙŖᝁĞ ༉⩶ᮥ ᯕᬊ⦹ᩍ ქᜅ᮹
ࠥ₊eᮥᩩ⊂⦹ᩡ݅. ქᜅ☖⧪eᨱ۵⦹₉eၰ႑₉ eĊᯕ Łಅࡹᨩ݅. ᯙŖᝁĞᮥ ᯕᬊ⦽ MAPE ᇥᕾđŝ ᧞
4~8%᮹᪅₉ᮉᮥᅕᩡᮝ໑, Historical data based models ၰ
⫭ȡ༉⩶ᨱ እ⧕ ׳ᮡ ᱶ⪶ࠥෝ ᅕᩡ݅.
Cherrett et al. (1996) ᩑǍᨱᕽ۵feed-forward ANN ༉⩶ᮥ
ᯕᬊ⦹ᩍยⓍ᮹☖⧪eᩩ⊂ᮥᝅ⦹ᩡ݅. ʑĥುᱢ(mechanistic)
ႊჶು᮹Ğᬑᱶℕᔢ⫊ᨱᕽ☖⧪eᩩ⊂ᨱᨕಅᬡᯕᯩᨩᮝ໑, ၹݡಽᯙŖᝁĞ᮹ĞᬑᮁŁၰᱶℕᔢ⫊ᨱᕽࠥ׳ᮡᩩ⊂ಆᮥ
ᅕᩡ݅.
Ohba et al. (1997) ᩑǍᨱᕽ۵ ☖⧪e ᩩ⊂ᨱ ᯩᨕ ⪝⧊
ᯙŖᝁĞ༉⩶ᮥᱽᦩ⦹ᩡ݅. ⪝⧊ᯙŖᝁĞ༉⩶ᮡᰆʑe
ၽᔾࡹ۵ᮁŁၰᱶℕᔢ⫊ᨱᱢ⧊⦽ᩩ⊂༉⩶ᮝಽᝅᱽŁᗮࠥಽ
ᯱഭෝ ☖⦽ ᇥᕾđŝ ׳ᮡ ᩩ⊂ಆᮥ ᅕᩡ݅.
2.4 টෘ֜ଭఙন
ᖁ⧪ᩑǍá☁ᨱᕽ⪶ᯙ⧁ᙹᯩॐᯕ⦽ǎŁᗮࠥಽ✚ᖒᨱ฿۵
ᩩ⊂ႊჶᨱݡ⦽ᝅ᷾ᩑǍ۵ᇡ᳒⦽ᝅᱶᯕ݅. ǎ᮹Ğᬑprove vehicles ᯱഭෝ≉ा⧁ᙹᯩ۵Ǎeᯕᱽ⦽ᱢᯕ໑, sample ᙹa
ᱢᨕᝅᱽᇥᕾᨱ⪽ᬊ⦹ʑᨱ۵⦽ĥaᯩ݅. ✚⯩ၙǎ᮹Mobile Millenium Project (2008)ᨱᕽᙹᯩॐᯕᝅᱽ☖⧪eᔑᱶ᮹
ᨕಅᬡᮥɚᅖ⦹ʑ᭥⧕GPSෝᰆ₊⦽ᜅษ✙⡑ᮥᯕᬊ⦹ᩍᇥᕾ Ǎeᨱ ݡ⦽ ᯱഭෝ ᙹḲ⦹ᩡ݅. ၹݡಽ ⦽ǎŁᗮࠥಽ᮹ Ğᬑ
Hi-pass ᯦ࠥ ⬥ probe vehiclesᯕ᷾aࡹᨕ ᝁᖒ ᯩ۵ᝅᱽ
☖⧪e đŝࠥ⇽ᯕ a܆⦹ᩍ Ʊ☖ඹ ᇥᕾ ၰ ᬕᩢᨱ ⪽ᬊᯕ
a܆⦹݅. ᯕᨱᅙᩑǍᨱᕽ۵ᝅᱽ☖⧪eᮥʑၹᮝಽ⦽DSRC
ᯱഭෝ ၵ┶ᮝಽ ᯙŖᝁĞᮥ ᯕᬊ⦽ ᩩ⊂ႊჶುᮥ ࠥ⇽⦽݅.
ᯕᨕᕽ ߅ӽᙹෝ ᯕᬊ⦽ ☖⧪e ᩩ⊂ಆ á᷾ᮥ ☖⧕ ʑ᳕
ᩑǍ᮹ ⦽ĥෝ ᪥⪵}ᖁ⦹Łᯱ ⦽݅.
3. ᩑǍႊჶು
3.1 ਓਏԩധෘਏԩુ౸
ᅙᩑǍᨱᱢᬊࡽႊჶುᮡDSRC ➉▕DBෝᯕᬊ⦽ᯙŖᝁĞ
ʑჶᮝಽFig. 1ŝz݅. ☖⧪e➉▕DB ၰᯙŖᝁĞᩩ⊂ŝ ᱶ ॒ᮝಽ Ⓧí ӹ٭ ᙹ ᯩᮝ໑, ᖙᇡ ŝᱶᮡ ݅ᮭŝ z݅.
Fig. 1. Structure of Methodology
[݉ĥ1] DSRC ᯱഭ᮹5ᇥḲĥ➉▕ᱶᅕෝǍ⇶⦹໑, đ⊂s
ၰ ᪅ඹ⊹ ᱽÑ
[݉ĥ2] ᯕಆᯱഭ᮹ Ğᬑ, 5 step eeĊ᮹ DSRC ĥᩕ
ᯱഭ ⠪⪽⪵ ᯲ᨦ(smoothing process)ᮥ ᝅ
ćÒ Î Ƈ áàÏ ā Ï Ƈ Ɔ ì ſƊƊ ƑſƋƎƊƃ
(1) Ɔ é ¡ƇƑƒƍƐƇƁſƊ ƂſƒſÞ¬« ƐſƔƃƊ ƇƋƃß
[݉ĥ3] ᝅe ᯱഭ᪡ ᯕಆᯱഭ᮹ ๅ⋎ᨱ۵ ᮁⓕญऽ Ñญ (Euclidean distance)ෝ ᔍᬊ
¦Ƈƌ ö ć ā Ƈ á Î ƌ Þ Ƈ Ɓ à Ƈ Ɔ ß Ï ì ſƊƊ ƑſƋƎƊƃ
(2) Ɓ é ƓƐƐƃƌƒ ƂſƒſÞ¬« ƐſƔƃƊ ƇƋƃß
(a) Historical data (b) Actual data Fig. 2. Pattern Matching by Euclidean Distance
[݉ĥ4] ᯕಆᯱഭ᮹༊⢽s(target)ᨱෙᯙŖᝁĞ⦺ၰ
aᵲ⊹ ᔑᱶ
Fig. 3. Process of ANN Learning and Weight
ⴗ aᵲ⊹
Ɣì ƕ
Ⅹʑ⪵ ၰ ⦺⬀ᙹƉ
ᖅᱶ ⴘ ⦺ᮉ(Ľ
) ၰ ↽ݡ᪅₉(Ƨ
) ᖅᱶ ⴙ ᮡܪ⊖ ĥᔑ§ Ƙ á ±¯
³ á ƄÞ§ Ƙ ß á ć Î âƃ à§
ƘÎ
ⴚ ⇽ಆ⊖ ĥᔑ
F ćÏ Î ÞƂàƗß Ï â
ⴛ ⇽ಆ⊖᮹ ᪅₉ᝁ⪙ ĥᔑ
ĺ Ɨ á ÞƂàƗßƗÞÎ à Ɨß
ⴜ ᮡܪ⊖᮹ ᪅₉ᝁ⪙ ĥᔑĺ Ƙ á ƘÞÎ à Ƙß ā Ƈ á Î Ƌ ĺ Ɨ ƕ
ⴝ aᵲ⊹ ᨦߑᯕ✙
ƕ á ƕ âķĺ Ɨ ³ Ɣ á Ɣ âķĺ Ƙ ±
ⴞ⨩ᬊ᪅₉ჵ᭥ᯝĞᬑ᳦ഭ, ⨩ᬊ᪅₉ჸᨕԁĞᬑⴙᮝಽ
ᯕ࠺
[݉ĥ5] ↽᳦ᩩ⊂⊹᮹᪅₉ᨱᩢ⨆ᮥᵝ۵᯦ಆ⊖ŝᮡܪ⊖᮹
ᙹ۵ ⧪₊᪅(trial and error) ႊჶᮥ ☖⧕ ᖁᱶ
[݉ĥ6] ᝅeᯱഭᨱʑၹ⦹ᩍࠥ⇽ࡽaᵲ⊹ᨱ᮹⧕☖⧪e
ᩩ⊂ đŝa ࠥ⇽
Fig. 4. Process of Travel Time Prediction
3.2 ധෘਏԩુ౸էրՑஹ
ᯙŖᝁĞᮥᯕᬊ⦽☖⧪eᩩ⊂đŝ᮹á᷾ᮥ᭥⧕ᕽ߅ ӽᙹෝᯕᬊ⧕☖⧪eᩩ⊂đŝ᮹ᱶ⪶ࠥෝá᷾⦽݅. ᯕෝ᭥⧕
ᕽᯕಆᯱഭ᮹⠪Ɂ, ⢽ᵡ⠙₉, ↽ᗭs, ↽ݡs॒᮹☖ĥ⊹ෝŁಅ
⦹ᩍ߅ӽᙹෝၽᔾ┉݅. ੱ⦽☖⧪eᩩ⊂đŝ᮹➉▕DB ᮹᳕ࠥෝእƱ⦹ʑ᭥⧕ᕽᝅeᯱഭᨱ ߅➉▕ᮥ⡍⧉⦹ᩍ
እƱᇥᕾᮥ ᝅ⦽݅.
4. ႊჶುᱢᬊၰᇥᕾ
4.1 ंজ֜ԩࢫୀ߹
ᅙᩑǍᨱᕽ۵⦽ǎŁᗮࠥಽᨱᕽᱽŖ⦹۵ĞᇡŁᗮࠥಽDSRC
ᯱഭෝᯕᬊ⦽݅. ᝅeᯱഭ۵5ᬵ6ᯝ(ɩ)ᯕ໑, ᯕಆᯱഭ۵ᝅ
Fig. 5. Study Site
Fig. 6. Optimum Input and Hidden Layer
Fig. 7. Prediction of Actual Travel Time
Fig. 8. Actual Predicted Result
Fig. 9. Actual Predicted Scatter Diagram
eᯱഭෝᱽ⦽2011֥3ᬵ1ᯝ~5ᬵ31ᯝᯱഭᯕ݅. ᇥᕾǍeᮡFig. 5᪡zᯕᦩᖒJC~᪅ᔑIC Ǎe᮹᳦ᱱႊ⨆(ᇡᔑᱱ)ᮥᖁᱶ
⦹ᩡ݅.
4.2 ਓਏԩധෘਏԩુ౸էր
ᅙᩑǍᨱᕽ۵ᯙŖᝁĞʑჶ᮹ᩩ⊂᪅₉ᨱᩢ⨆ᮥᵝ۵᯦ಆ
⊖ŝᮡܪ⊖᮹ᙹ۵⧪₊᪅(trial and error) ႊჶᮥᯕᬊ⦹ᩡ݅.
ᇥᕾđŝ Fig. 6ŝ zᯕ ᯦ಆ⊖ 3}, ᮡܪ⊖ 5}ᨱᕽ ᪅₉ᮉᯕ
aᰆ ԏᦹᮝ໑, ᯦ಆ⊖ ၰ ᮡܪ⊖᮹ ᙹa ۹ᨕԁᙹಾ ᪅₉ᮉᯕ
᷾a⦹۵ Ğ⨆ᮥ ᅕᩡ݅.
ᯕᨕᕽᯙŖᝁĞᮥᯕᬊ⦽☖⧪eᩩ⊂ᮥ᭥⧕ᩩ⊂ᵝʑෝ
ෙᇥᕾᮥᝅ⦹ᩡ݅. ຝᱡFig. 7ŝzᯕᝅeDSRC ʑၹ
☖⧪eᩩ⊂ᮥᝅ⦹ᩡ݅. ᇥᕾđŝᝅ⊂☖⧪eŝ᮹᪅₉ᮉ ᯕ3.95%ಽᬑᙹ⦽ᩩ⊂ಆᮥᅕᩡ݅. ੱ⦽ᔑᱱࠥᇥᕾđŝR
2
ᯕ0.96ᮝಽ ׳ᮡ ᔢššĥෝ ᅕᩡ݅.
4.3 ധෘਏԩુ౸էրՑஹէր
ᯙŖᝁĞᮥᯕᬊ⦽☖⧪eᩩ⊂đŝ᮹á᷾ᮥ᭥⧕ᕽ߅ ӽᙹෝᯕᬊ⦽☖⧪eᩩ⊂đŝ᪡ᱶ⪶ࠥෝእƱ⦹ᩡ݅. ᯕෝ
᭥⧕ᕽ߅ӽᙹၽᔾᯕಆᯱഭ᮹⠪Ɂ, ⢽ᵡ⠙₉, ↽ᗭs, ↽ݡ s॒᮹☖ĥ⊹ෝŁಅ⦹ᩡᮝ໑, Table 1ŝz݅. ੱ⦽☖⧪e
ᩩ⊂đŝ᮹ ➉▕DB ᮹᳕ࠥෝ እƱ⦹ʑ ᭥⧕ᕽ ᝅe ᯱഭᨱ
Table 1. Statistics of DSRC
Classification Historical data Actual data
Average 7.97 8.07
Standard deviation 3.46 3.31
Min. 5.62 5.29
Max. 35.89 20.76
Fig. 10. Optimum Input and Hidden Layer (Random Number)
Fig. 12. Actual Predicted Scatter Diagram (Perfect Matching)
Fig. 13. Actual Predicted Result (Random Number)
߅➉▕ᮥ ⡍⧉⦹ᩍ እƱᇥᕾᮥ ᝅ⦹ᩡ݅.
ᯙŖᝁĞ᮹ᩩ⊂᪅₉ᨱᩢ⨆ᮥᵝ۵᯦ಆ⊖ŝᮡܪ⊖᮹ᙹ۵
⧪₊᪅ŝᱶᮥ☖⧕Fig. 10ŝ zᯕ᯦ಆ⊖9}, ᮡܪ⊖9}ಽ
ࠥ⇽ࡹᨩ݅. ᯙŖᝁĞ᮹☖⧪eᩩ⊂đŝᨱݡ⦽➉▕᮹᳕ࠥ
۵Figs. 11~14᮹đŝ᪡z݅. ☖⧪eᯕಆᯱഭ᮹➉▕ᨱ߅ӽ ᙹෝ⡍⧉⦽Ğᬑᝅ⊂☖⧪eŝ᮹᪅₉ᮉᯕ0.0046%ಽ׳ᮡ
ᱶ⪶ࠥෝᅕᩡ݅. ੱ⦽ᔑᱱࠥᇥᕾđŝR
2
ᯕ1ಽ׳ᮡᔢššĥෝᅕᩡ݅.
ၹݡಽ߅ӽᙹaᯕಆᯱഭ᮹➉▕ᨱ⡍⧉ࡹḡᦫᮡĞᬑෝ
ᔕ⠕ᅕ໕, ☖⧪eᩩ⊂đŝ᪅₉ᮉᯕ18.98%ಽ׳íࠥ⇽ࡹᨩ
݅. ᯕ۵ᯙŖᝁĞᮥᯕᬊ⦽☖⧪eᩩ⊂➉▕DBaᩩ⊂᮹
ᱶ⪶ࠥᨱᵝ⦹í᯲ᬊ⦽đŝಽ❱݉ࡽ݅. ݅ั⧕, ☖⧪e
ᯕಆᯱഭ᮹➉▕ᨱ߅ӽᙹa⡍⧉ࡹḡ༜⧕ ᇡᱶ⪶⦽đŝෝ
ࠥ⇽⦹ᩡ݅. ੱ⦽ᔑᱱࠥᇥᕾđŝR
2
ᯕ0.51ಽԏᮡᔢššĥෝᅕᩡ݅.
5. đು
5.1 էߨ
ᅙᩑǍᨱᕽ۵ᝅᱽ☖⧪eᮥʑၹᮝಽ⦽DSRC ᯱഭෝၵ┶
ᮝಽ⦽ǎŁᗮࠥಽᨱᱢᱩ⦽ᩩ⊂ႊჶುᮥࠥ⇽⦹ᩡ݅. ĞᇡŁᗮ
ࠥಽᦩᖒJC~᪅ᔑIC Ǎe᮹24eDSRC ᯱഭෝᯕᬊ⦹໑݉ᵝ ʑ☖⧪eᩩ⊂ၰእᖁ⩶šĥᨱᕽ׳ᮡᱶ⪶ࠥෝᅕᯕ۵ᯙŖᝁ Ğʑჶᮥᱢᬊ⦹ᩡ݅. ੱ⦽߅ӽᙹෝᯕᬊ⦽☖⧪eᩩ⊂đ ŝ᮹ᱶ⪶ࠥá᷾ᮥ☖⧕ᮁᬊᖒá᷾ᮥᝅ⦹ᩡ݅. ᝅeDSRC ʑၹ☖⧪eᩩ⊂đŝ᪅₉ᮉᯕ3.95%ಽᬑᙹ⦽ᩩ⊂ಆᮥᅕᩡ
݅. ᯕ۵➉▕ʑၹ᮹ᯙŖᝁĞᩩ⊂ᯕಆᯱഭ᮹ᱥญŝᱶŝ
↽ᱢ᮹᯦ಆ⊖ၰᮡܪ⊖᮹ᖁᱶᮝಽᯙ⦽đŝಽ❱݉ࡽ݅. ☖⧪
eᩩ⊂đŝ᮹á᷾ᮥ᭥⧕ᕽ߅ӽᙹෝၽᔾ⦹ᩍእƱᇥᕾᮥ
ᝅ⦹ᩡ݅. ᇥᕾđŝᝅ⊂⊹᪡᮹᪅₉ᮉᯕ18.98%ಽ׳íࠥ⇽ࡹ
ᨩᮝ໑, ᯕ۵ ᯙŖᝁĞᮥ ᯕᬊ⦽ ☖⧪e ᩩ⊂ ➉▕DBa
ᩩ⊂᮹ ᱶ⪶ࠥᨱ ᵝ⦹í ᯲ᬊ⦽ đŝಽ ❱݉ࡽ݅.
ᖁ⧪ᩑǍá☁ᨱᕽ⪶ᯙ⧁ᙹᯩॐᯕ⦽ǎŁᗮࠥಽ✚ᖒᨱ฿۵
ᩩ⊂ႊჶၰeeĊᨱݡ⦽ᝅ᷾ᩑǍ۵ᇡ᳒⦽ᝅᱶᯕ݅. ⦹ḡอ
ᅙᩑǍ᮹ᝅ᷾ᇥᕾđŝෝᯕᬊ⦹ᩍ⦽ǎŁᗮࠥಽ✚ᖒᨱ฿۵
ᩩ⊂݉᭥ၰᱶᅕᱽŖᵝʑᨱݡ⦽ᖁᱶᯕa܆⧁äᮝಽ❱݉ࡽ݅.
ੱ⦽ ᝅᱽ Ʊ☖ᯱഭෝ ᯕᬊ⦽ ☖⧪e ᩩ⊂ᨱ ᯩᨕ ᯥ᮹ᖒᮥ
႑ᱽ⧁ ᙹ ᯩᨕ ᦩᱶᱢᯙ ᩑǍđŝ ⠪a⧁ ᙹ ᯩᮥ äᯕ݅.
5.2 ේบ֜ր୪
ᅙᩑǍ᮹⦽ĥᱱᮝಽ۵݉ᵝʑ☖⧪eᩩ⊂ၰእᖁ⩶šĥᨱ ᕽ׳ᮡᱶ⪶ࠥෝᅕᯕ۵ᯙŖᝁĞʑჶᮥ⪽ᬊ⦹ᩡ݅. ☖ĥᱢ
༉⩶᮹Ğᬑ☖⧪eᩩ⊂᮹ᱶ⪶ࠥ۵ᬑᙹ⧁ᙹᯩḡอ, ᩩ⊂đŝ ᨱݡ⦽ᯙŝšĥᇥᕾᨱᯩᨕဍ౩ᯕᖹ༉⩶ᨱእ⧕ᨕಅᬡᯕ
ᯩ݅. ᯕᨱ⨆⬥ᩑǍŝᱽಽဍ౩ᯕᖹ༉⩶ᮥᯕᬊ⦽ᩩ⊂đŝෝ
ၵ┶ᮝಽ☖ĥᱢ༉⩶ŝ᮹ᱶ⪶ࠥၰᯙŝšĥᇥᕾᯕ⦥⦹݅.
qᔍ᮹ɡ
ᅙᩑǍ۵ǎ☁⧕᧲ᇡ℉݉ࠥ}ၽᩑǍ}ၽᔍᨦ᮹ᩑǍእḡ ᬱ(12Ʊ☖ℕĥ-ḡ܆01)ᨱ ᮹⧕ ᙹ⧪ࡹᨩܩ݅.
