KSC 2018
Current Result Document :
ÇѱÛÁ¦¸ñ(Korean Title) |
ÀÚÀ²ÁÖÇàÀÚµ¿Â÷ÀÇ ¾ÈÀü ÁÖÇàÀ» À§ÇÑ °ÈÇнÀ ¾Ë°í¸®ÁòÀÇ ¿¬¼ÓÀûÀÎ Çൿ °ø°£ ÀÌ»êÈ |
¿µ¹®Á¦¸ñ(English Title) |
Discretization of continuous Action Space of Reinforcement Algorithm for Safe Autonomous Vehicle Driving |
ÀúÀÚ(Author) |
ÁÖÈ£ÅÃ
¹ÚÇöö
±è°æÁß
HoTaek Joo
HyunCheol Park
Kyungjoong Kim
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 45 NO. 02 PP. 0776 ~ 0778 (2018. 12) |
Çѱ۳»¿ë (Korean Abstract) |
ÀÚÀ²ÁÖÇà ÀÚµ¿Â÷ ºÐ¾ß¿¡¼ °ÈÇнÀ(Reinforcement Learning) Àº µö·¯´× (deep learning)°ú ÇÔ²² °¡Àå ÁÖ¸ñ¹Þ´Â ±â¼úÀÌ´Ù. °ÈÇнÀÀ̶õ ±â°è ÇнÀÀÇ ¹æ¹ý Áß Çϳª·Î ÇöÀç »óÅÂ(state)¿¡¼ ¾î¶² Çൿ(action)À» ¼±ÅÃÇßÀ» ¶§ ¹Þ´Â ¹Ì·¡ º¸»ó(reward)ÀÇ ÈûÀ» ÃÖ´ëÈÇÏ´Â ¹æÇâÀ¸·Î ÇൿÇÏ¿© ÇнÀÇÏ´Â ¹æ¹ýÀÌ´Ù. º» ³í¹®ÀÇ ¼±Çà ¿¬±¸¿¡¼ ÀÚÀ²ÁÖÇà ÀÚµ¿Â÷¿Í °°Àº ¿¬¼ÓÀûÀÎ Çൿ °ø°£(Continuous Action Space)¿¡¼ ±âÁ¸ÀÇ °ÈÇнÀÀ» Àû¿ëÇϸé Á÷¼±ÄÚ½º¿¡¼ÀÇ ÇÚµé ¶³¸² ¹®Á¦¸¦ È®ÀÎÇÏ¿´´Ù. º» ³í¹®¿¡¼´Â ÀÌ ¹®Á¦¸¦ ÇØ°áÇϱâ À§ÇÏ¿© ¿¬¼ÓÀûÀÎ Çൿ °ø°£À» ÀÌ»êÈ(discretization)ÇÏ´Â ¹æ¹ýÀ» Á¦½ÃÇÏ¿´À¸¸ç, À̸¦ ±¸ÇöÇϱâ À§ÇØ °æÁÖ¿ë Â÷·® ½Ã¹Ä·¹À̼ÇÀÎ TORCS(The Open Racing Car Simulator)¿¡¼ °ÈÇнÀ ¾Ë°í¸®Áò Áß ÇϳªÀÎ PPO(Proximal Policy Optimization)À» Á¢¸ñÇÏ¿© ½ÇÇèÀ» ÁøÇàÇÏ¿´´Ù. |
¿µ¹®³»¿ë (English Abstract) |
|
Å°¿öµå(Keyword) |
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|