• ๋Œ€ํ•œ์ „๊ธฐํ•™ํšŒ
Mobile QR Code QR CODE : The Transactions of the Korean Institute of Electrical Engineers
  • COPE
  • kcse
  • ํ•œ๊ตญ๊ณผํ•™๊ธฐ์ˆ ๋‹จ์ฒด์ด์—ฐํ•ฉํšŒ
  • ํ•œ๊ตญํ•™์ˆ ์ง€์ธ์šฉ์ƒ‰์ธ
  • Scopus
  • crossref
  • orcid

  1. (School of Electronics Engineering, Kyungpook National University, Korea)



Multi-sensor system, MPC, Autonomous mobile robot, State estimator, Tracking control

1. ์„œ๋ก 

์ž์œจ ์ฃผํ–‰ ์‹œ์Šคํ…œ์˜ ๊ฐœ๋ฐœ์„ ์œ„ํ•˜์—ฌ ๋‹ค์–‘ํ•œ ์„ผ์„œ๋“ค์„ ์œตํ•ฉํ•˜์—ฌ ์‚ฌ์šฉํ•˜๋Š” ๋ฐฉ๋ฒ•๋“ค์ด ์‹œ๋„๋˜๊ณ  ์žˆ๋‹ค[1-3]. ํŠนํžˆ ์ง€๋Šฅ ์‹œ์Šคํ…œ์˜ ๊ตฌํ˜„์„ ์œ„ํ•˜์—ฌ ์นด๋ฉ”๋ผ์™€ ๋ผ์ด๋” ๊ฐ™์€ ์ฃผ๋ณ€์œผ๋กœ๋ถ€ํ„ฐ ๋งŽ์€ ์–‘์˜ ์ •๋ณด๋ฅผ ํš๋“ํ•  ์ˆ˜ ์žˆ๋Š” ์„ผ์„œ๋“ค์ด ๋งŽ์ด ์ด์šฉ๋˜๊ณ  ์žˆ๋‹ค[4]. ์ž์œจ ์ฃผํ–‰ ์‹œ์Šคํ…œ์˜ ์œ„์น˜ ๋ณ€ํ™”๋ฅผ ์ธก์ •ํ•˜๊ธฐ ์œ„ํ•œ ๊ฒฝ์šฐ์—๋„ ์นด๋ฉ”๋ผ, ๋ผ์ด๋‹ค, ์—”์ฝ”๋”, ๊ด€์„ฑ ์ธก์ • ์žฅ์น˜ (IMU) ๋“ฑ๊ณผ ๊ฐ™์€ ๋‹ค์–‘ํ•œ ์„ผ์„œ๋“ค์ด ์‚ฌ์šฉ๋œ๋‹ค. ๊ฐ๊ฐ์˜ ์„ผ์„œ๋“ค์€ ์žฅ๋‹จ์ ์„ ๊ฐ€์ง€๊ณ  ์žˆ๋Š”๋ฐ, ๋ผ์ด๋‹ค๋Š” ๋น›์„ ๋ฐœ์ƒ์‹œํ‚ค๋ฉด์„œ ์‚ฌ์šฉ๋˜๋Š”๋ฐ, ์ด๋Š” ๋ Œ์ฆˆ์˜ ์ƒํƒœ์™€ ๋ฐ˜์‚ฌํ•˜๋Š” ๋ชฉํ‘œ๋ฌผ์˜ ์žฌ์งˆ์— ๋”ฐ๋ผ์„œ ์„ฑ๋Šฅ์ด ๋‹ฌ๋ผ์ง„๋‹ค. ์—”์ฝ”๋”๋Š” ์‹œ์Šคํ…œ์— ํ†ตํ•ฉ๋˜๊ธฐ ํŽธ๋ฆฌํ•˜๊ณ , ์‹ค์‹œ๊ฐ„์œผ๋กœ ์ •๋ณด๋ฅผ ์ œ๊ณตํ•œ๋‹ค๋Š” ์žฅ์ ์ด ์žˆ์ง€๋งŒ, ์™ธ๋ถ€ํ™˜๊ฒฝ์— ๋”ฐ๋ผ ํฌ๊ฒŒ ์ •๋ณด์˜ ํ™•์‹ค์„ฑ์ด ๋–จ์–ด์ง€๋Š” ๋‹จ์ ์ด ์žˆ๋‹ค. ์นด๋ฉ”๋ผ๋Š” ๋ฌผ๋ฆฌ์ ์ธ ํŠน์ง•๋“ค์„ ๊ฐ€์ง€๊ณ  ๋ฌผ์ฒด๋“ค์„ ๊ตฌ๋ณ„ํ•˜๊ฑฐ๋‚˜ ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ํ†ตํ•˜์—ฌ ์ •ํ™•ํ•œ ์ •๋ณด๋ฅผ ์–ป์„ ์ˆ˜ ์žˆ๋‹ค. ํ•˜์ง€๋งŒ, ๋ณต์žกํ•œ ์‹ ํ˜ธ ์ฒ˜๋ฆฌ ๊ณผ์ •๋“ค ๋•Œ๋ฌธ์—, ์‹ค์‹œ๊ฐ„์œผ๋กœ ์ •๋ณด๋ฅผ ๋ฐ›๊ธฐ๊ฐ€ ์–ด๋ ต๋‹ค. ์ด๋Ÿฌํ•œ ์„ผ์„œ๋“ค์˜ ๊ฐ๊ฐ์˜ ์žฅ๋‹จ์ ์„ ๊ทน๋ณตํ•˜๊ธฐ ์œ„ํ•ด์„œ ์„ผ์„œ๋“ค์„ ์œตํ•ฉํ•˜์—ฌ ์‹œ์Šคํ…œ์„ ๊ตฌ์„ฑํ•˜๋Š” ์ ‘๊ทผ๋ฒ•์ด ์—ฐ๊ตฌ๋˜๊ณ  ์žˆ๋‹ค[5]. ๋‹ค์–‘ํ•œ ์„ผ์„œ๋ฅผ ํ†ตํ•œ ์œ„์น˜ ์ธก์ • ๋ฐฉ๋ฒ•์œผ๋กœ ๊ธ€๋กœ๋ฒŒ ์œ„์„ฑ ํ•ญ๋ฒ• ์‹œ์Šคํ…œ, ๋ผ์ด๋‹ค, ๊ด€์„ฑ ์ธก์ • ์žฅ์น˜๋ฅผ ์ด์šฉํ•œ ๋ฐฉ๋ฒ•์ด ๊ฐœ๋ฐœ๋˜์–ด ์ •ํ™•๋„๋Š” ํ–ฅ์ƒ๋˜์—ˆ์ง€๋งŒ ์ด๋Ÿฌํ•œ ๋ฐฉ๋ฒ•์€ ๋‹จ์ˆœํ•˜๊ฒŒ ์ƒ˜ํ”Œ๋ง ์‹œ๊ฐ„์ด ๊ฐ€์žฅ ๋น ๋ฅธ ๋ฐ์ดํ„ฐ์— ๋™๊ธฐํ™”ํ•˜์—ฌ ์‚ฌ์šฉํ•˜๋Š” ๋ฐฉ์‹์œผ๋กœ ์ƒ˜ํ”Œ๋ง ๋น„๋™๊ธฐํ™”์— ๋Œ€ํ•œ ๋ฌธ์ œ๋Š” ๋‹ค๋ฃจ์ง€ ๋ชปํ–ˆ๋‹ค[6]. ์ด๋Ÿฌํ•œ ๋ฌธ์ œ๋ฅผ ํ•ด๊ฒฐํ•˜๊ธฐ ์œ„ํ•ด ์นผ๋งŒ ํ•„ํ„ฐ๋ฅผ ์ด์šฉํ•˜์—ฌ ์„ผ์„œ ๋ฐ์ดํ„ฐ๋ฅผ ์ฒ˜๋ฆฌํ•˜๋Š” ๋ฐฉ๋ฒ•์ด ๋งŽ์ด ์ด์šฉ๋˜๊ณ  ์žˆ๋‹ค[7]. ๊ทธ๋Ÿฌ๋‚˜ ์นผ๋งŒ ํ•„ํ„ฐ๋ฅผ ์ด์šฉํ•˜๋Š” ์ ‘๊ทผ๋ฒ• ์—ญ์‹œ ๊ฐ๊ฐ์˜ ์„ผ์„œ๋“ค์˜ ์ƒ˜ํ”Œ๋ง ์ฃผ๊ธฐ๋“ค์„ ๊ณ ๋ คํ•˜์—ฌ ์ƒํƒœ ์ถ”์ •๊ธฐ์˜ ์•ˆ์ •์„ฑ์„ ๋ณด์žฅํ•˜๋Š” ์—ฐ๊ตฌ๋Š” ๋ฏธํกํ•œ ์‹ค์ •์ด๋‹ค. ๋”ฐ๋ผ์„œ ๋ณธ ๋…ผ๋ฌธ์—์„œ๋Š” ๋‹ค์ค‘ ์ƒ˜ํ”Œ๋ง ์ฃผ๊ธฐ๋ฅผ ๊ฐ€์ง€๋Š” ์—ฌ๋Ÿฌ ๊ฐ€์ง€ ์„ผ์„œ์—์„œ ์–ป์–ด์ง€๋Š” ์ •๋ณด๋กœ๋ถ€ํ„ฐ ์‹œ์Šคํ…œ ์ƒํƒœ๋ฅผ ์ถ”์ •ํ•˜๋Š” ๋‹ค์ค‘ ์ƒ˜ํ”Œ๋ฐ์ดํƒ€ ๊ธฐ๋ฐ˜์˜ ์ƒํƒœ์ถ”์ •๊ธฐ์™€ ์ถ”์ ์ œ์–ด๋ฅผ ์œ„ํ•œ ์—๋Ÿฌ๋ชจ๋ธ ๊ธฐ๋ฐ˜์˜ ์ตœ์ ํ™”๊ฐ€ ๊ฐ€๋Šฅํ•œ ๋ชจ๋ธ์˜ˆ์ธก์ œ์–ด๊ธฐ๋กœ ๊ตฌ์„ฑ๋œ ์ด๋™๋กœ๋ด‡์„ ์œ„ํ•œ ์ถ”์ ์ œ์–ด์‹œ์Šคํ…œ์˜ ์ƒˆ๋กœ์šด ๊ตฌ์กฐ์™€ ๊ตฌํ˜„ ๋ฐฉ๋ฒ•์„ ์ œ์‹œํ•œ๋‹ค.

๋ชจ๋ธ ์˜ˆ์ธก ์ œ์–ด ๊ธฐ๋ฒ•์€ ๋ชจ๋ฐ”์ผ ๋กœ๋ด‡์„ ์ œ์–ดํ•˜๊ธฐ ์œ„ํ•˜์—ฌ ๋งŽ์ด ์‚ฌ์šฉ๋˜๊ณ  ์žˆ๋‹ค[8]. ์‹œ์Šคํ…œ์˜ ๋ชจ๋ธ์„ ๊ธฐ๋ฐ˜์œผ๋กœ ๋ฏธ๋ž˜์˜ ์ถœ๋ ฅ์น˜๋ฅผ ๋ฏธ๋ฆฌ ๊ณ„์‚ฐํ•˜์—ฌ ์˜ˆ์ธกํ•˜๊ณ  ์ด๋ฅผ ์ตœ์ ํ™”ํ•˜๊ธฐ ์œ„ํ•˜์—ฌ ์–ป์–ด์ง„ ์ž…๋ ฅ์„ ์‚ฌ์šฉํ•จ์œผ๋กœ ์ œ์–ด ์„ฑ๋Šฅ์„ ๋†’์ผ ์ˆ˜ ์žˆ๋‹ค๋Š” ์žฅ์ ์ด ์žˆ๋‹ค. ๋˜ํ•œ ์ด๋™ ๋กœ๋ด‡์˜ ์ œํ•œ ์กฐ๊ฑด์ด๋‚˜ ๋กœ๋ด‡์˜ ์ž„๋ฌด ์ˆ˜ํ–‰ ์ค‘์— ๋ฐœ์ƒํ•˜๋Š” ์ œํ•œ ์กฐ๊ฑด์„ ์ตœ์ ํ™” ๊ณผ์ •์— ์ ์šฉํ•  ์ˆ˜ ์žˆ์–ด ์‹ค์ œ ์‹œ์Šคํ…œ์— ์ ์šฉํ•˜๋Š” ๋ฌธ์ œ์— ๋งŽ์€ ์ด์ ์„ ์ œ๊ณตํ•œ๋‹ค. ๊ทธ๋Ÿฌ๋‚˜ ๋ชจ๋ธ ์˜ˆ์ธก ์ œ์–ด๋Š” ๋งŽ์€ ๊ณ„์‚ฐ๋Ÿ‰์ด ํ•„์š”ํ•œ ์ œ์–ด ๊ธฐ๋ฒ•์œผ๋กœ ์ž„๋ฒ ๋””๋“œ ์‹œ์Šคํ…œ์— ์ ์šฉํ•˜๋Š” ๊ฒฝ์šฐ์—๋Š” ๋งŽ์€ ๋ฌธ์ œ์ ๋“ค์ด ๋ฐœ์ƒํ•œ๋‹ค. ํŠนํžˆ ์ด๋™ ๋กœ๋ด‡๊ณผ ๊ฐ™์ด ์ด๋™ํ•˜๋Š” ์‹œ์Šคํ…œ์˜ ๊ฒฝ์šฐ ์ œ์–ด ์ž…๋ ฅ์˜ ์ง€์—ฐ์œผ๋กœ ์ถฉ๋Œ์ด๋‚˜ ์ œ์–ด ๋ถˆ๋Šฅ ์ƒํƒœ๊ฐ€ ๋ฐœ์ƒํ•  ์ˆ˜ ์žˆ๋‹ค.

๋ณธ ์—ฐ๊ตฌ์—์„œ๋Š” ๋‹ค์ค‘ ์„ผ์„œ์˜ ์œตํ•ฉ ๋ฌธ์ œ์™€ ๋ชจ๋ธ ์˜ˆ์ธก ์ œ์–ด์˜ ์ž…๋ ฅ ๊ณ„์‚ฐ ์‹œ๊ฐ„ ์ง€์—ฐ ๋ฌธ์ œ๋ฅผ ํ•ด๊ฒฐํ•˜๊ธฐ ์œ„ํ•˜์—ฌ ๋‹ค์ค‘ ์„ผ์„œ๋ฅผ ๊ฐ€์ง€๋Š” ์ƒํƒœ ์ถ”์ •๊ธฐ์™€ ์‹ค์‹œ๊ฐ„ ์ œ์–ด๋ฅผ ์œ„ํ•œ ์™ธ์—ฐ์  ๋ชจ๋ธ ์˜ˆ์ธก ์ œ์–ด๊ธฐ๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœํ•œ ์ด๋™ ๋กœ๋ด‡ ์ถ”์ ์ œ์–ด ์‹œ์Šคํ…œ์ด ์—ฐ๊ตฌ๋œ๋‹ค. ์ œ์•ˆ๋œ ์‹œ์Šคํ…œ์„ ๊ตฌํ˜„ํ•˜๊ธฐ ์œ„ํ•˜์—ฌ ๋กœ๋ด‡ ์šด์˜ ์ฒด์ œ๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ์ž„๋ฒ ๋””๋“œ ํ™˜๊ฒฝ์—์„œ ๊ฐœ๋ฐœ๋˜์—ˆ๋‹ค. ๋กœ๋ด‡ ์šด์˜ ์ฒด์ œ์— ๊ตฌํ˜„๋œ ์ž์œจ ์ฃผํ–‰ ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ํ†ตํ•ด ์นด๋ฉ”๋ผ, ์—”์ฝ”๋”, ๋ผ์ด๋”๋กœ๋ถ€ํ„ฐ ๋ฐ์ดํ„ฐ๋ฅผ ํš๋“ํ•˜๊ณ  ์ฃผ๋ณ€๊ณผ ๋กœ๋ด‡์˜ ์ƒํƒœ๋ฅผ ํŒ๋‹จํ•˜๊ธฐ ์œ„ํ•ด ๋ฐ์ดํ„ฐ๋“ค์„ ์ฒ˜๋ฆฌํ•œ๋‹ค. ๋˜ํ•œ ์–ป์–ด์ง„ ์ƒํƒœ ์ •๋ณด๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ๋กœ๋ด‡์˜ ์†๋„ ๋ฐ ์œ„์น˜๋ฅผ ์ œ์–ดํ•˜์—ฌ ์ž์œจ ์ด๋™๋กœ๋ด‡์˜ ์ถ”์  ์ œ์–ด์— ์ ์šฉ๋œ๋‹ค. ๋ณธ ๋…ผ๋ฌธ์˜ ๊ธฐ์—ฌ๋„๋Š” ๋‹ค์Œ๊ณผ ๊ฐ™์ด ์š”์•ฝ๋œ๋‹ค.

ยท ์ƒ˜ํ”Œ ๋ฐ์ดํ„ฐ ์ƒํƒœ ์ถ”์ •๊ธฐ ๊ธฐ๋ฐ˜์˜ ์™ธ์—ฐ์  ๋ชจ๋ธ ์˜ˆ์ธก์ œ์–ด ๊ตฌ์กฐ๊ฐ€ ์ œ์•ˆ๋˜์–ด์ง„๋‹ค.

ยท ๋‹ค๋ฅธ ์ฃผ๊ธฐ๋ฅผ ๊ฐ€์ง€๋Š” ์„ผ์„œ๋“ค์˜ ์‹ ํ˜ธ๋ฅผ ์œตํ•ฉํ•  ์ˆ˜ ์žˆ๋Š” ์ƒํƒœ ์ถ”์ •๊ธฐ๊ฐ€ ์„ค๊ณ„๋œ๋‹ค.

ยท ํ‘œ์ง€ํŒ ๊ฒ€์ถœ, ๋ผ์ด๋” ๊ธฐ๋ฐ˜์˜ ๊ฒฝ๋กœ ์ƒ์„ฑ ์•Œ๊ณ ๋ฆฌ์ฆ˜์ด ์ž์œจ ์ด๋™๋กœ๋ด‡์„ ์œ„ํ•ด ๊ตฌํ˜„๋˜๊ณ  ์ƒํƒœ ์ถ”์ •๊ธฐ, ๋ชจ๋ธ์˜ˆ์ธก๊ธฐ์™€ ํ•จ๊ป˜ ํ†ตํ•ฉ๋˜์–ด ์‹คํ—˜๋œ๋‹ค.

๋…ผ๋ฌธ์˜ ๋‚˜๋จธ์ง€ ๋ถ€๋ถ„์€ ๋‹ค์Œ๊ณผ ๊ฐ™์ด ๊ตฌ์„ฑ๋˜์–ด ์žˆ๋‹ค. 2์žฅ๊ณผ 3์žฅ์—์„œ๋Š” ์ œ์•ˆ๋œ ์‹œ์Šคํ…œ์˜ ๊ตฌ์„ฑ ๋ฐ ๊ตฌํ˜„์— ๋Œ€ํ•œ ์„ค๋ช…, 4์žฅ์—์„œ๋Š” ์ œ์•ˆ๋œ ๋ฐฉ๋ฒ•๋“ค์„ ์‹œ๋ฎฌ๋ ˆ์ด์…˜ ๋ฐ ์‹คํ—˜์— ์ ์šฉํ•˜์—ฌ ์–ป์–ด์ง„ ๊ฒฐ๊ณผ๋“ค์„ ์ •๋ฆฌํ•˜์˜€๋‹ค. 5์žฅ์—์„œ๋Š” ์—ฐ๊ตฌ๋œ ๊ฒฐ๊ณผ๋“ค์— ๋Œ€ํ•ด ๋…ผ์˜ํ•œ๋‹ค.

2. ์‹œ์Šคํ…œ ๊ตฌ์„ฑ

์ž์œจ ์ฃผํ–‰ ์‹œ์Šคํ…œ์˜ ๊ตฌํ˜„์„ ์œ„ํ•ด ๋ณธ ์—ฐ๊ตฌ์—์„œ๋Š” ๋ฌด์ธ ์ง€์ƒ ์ฐจ๋Ÿ‰ (HUSKY UGV), ์ž„๋ฒ ๋””๋“œ ๋ณด๋“œ (NVIDIA Jetson TX2), ์Šคํ…Œ๋ ˆ์˜ค ์นด๋ฉ”๋ผ (Zed camera), ๋ผ์ด๋” ์„ผ์„œ (Velodyne LiDAR)๋ฅผ ์ด์šฉํ•˜์˜€๋‹ค.

๋ฌด์ธ ์ง€์ƒ ์ฐจ๋Ÿ‰(HUSKY UGV)์—๋Š” ๋ชจํ„ฐ ์ปจํŠธ๋กค ๋ฐ ์—”์ฝ”๋” ๋ฐ์ดํ„ฐ ํš๋“์„ ์œ„ํ•œ ์ปจํŠธ๋กค ๋ณด๋“œ๊ฐ€ ํƒ‘์žฌ๋˜์–ด ์žˆ์–ด ์ž„๋ฒ ๋””๋“œ ๋ณด๋“œ (NVIDIA Jetson TX2)์™€ ์ง๋ ฌ ํ†ต์‹ ์„ ํ†ตํ•ด ์ œ์–ด ์ž…๋ ฅ๊ณผ ์„ผ์„œ ๋ฐ์ดํ„ฐ ๊ฐ’์„ ์ฃผ๊ณ ๋ฐ›๋Š”๋‹ค. ์ž„๋ฒ ๋””๋“œ ๋ณด๋“œ (NVIDIA Jetson TX2)์—๋Š” ์˜์ƒ ์ฒ˜๋ฆฌ๋ฅผ ์œ„ํ•œ ๊ทธ๋ž˜ํ”ฝ ์ฒ˜๋ฆฌ ์žฅ์น˜๊ฐ€ ๋‚ด์žฅ๋˜์–ด ์žˆ์–ด ์นด๋ฉ”๋ผ๋กœ ํš๋“ํ•œ ์ด๋ฏธ์ง€๋กœ๋ถ€ํ„ฐ ๋กœ๋ด‡์˜ ์œ„์น˜๋ฅผ ๊ณ„์‚ฐํ•  ์ˆ˜ ์žˆ๋‹ค. ์ฃผ์–ด์ง„ ํ™˜๊ฒฝ์—์„œ ๊ฒฝ๋กœ๋ฅผ ์ƒ์„ฑํ•˜๊ธฐ ์œ„ํ•ด์„œ ๋ผ์ด๋” ์„ผ์„œ๊ฐ€ ์ด์šฉ๋˜์—ˆ๋‹ค. ๊ทธ๋ฆผ. 1์€ ์ฃผ์–ด์ง„ ์‹œ์Šคํ…œ์˜ ๊ตฌ์„ฑ์„ ๋ณด์—ฌ์ค€๋‹ค. ์ž„๋ฒ ๋””๋“œ ๋ณด๋“œ (NVIDIA Jetson TX2)์—๋Š” ์ƒํƒœ ์ถ”์ •๊ธฐ, ์ œ์–ด๊ธฐ, ๊ฒฝ๋กœ ์ƒ์„ฑ ์•Œ๊ณ ๋ฆฌ์ฆ˜์ด ๊ตฌํ˜„๋˜์–ด ์žˆ์–ด, ๋ผ์ด๋‹ค์™€ ์นด๋ฉ”๋ผ๋ฅผ ํ†ตํ•ด ๊ฒฝ๋กœ๋ฅผ ์ƒ์„ฑํ•˜๊ณ  ์นด๋ฉ”๋ผ์™€ ์—”์ฝ”๋”๋กœ๋ถ€ํ„ฐ ์„ผ์„œ ๊ฐ’์„ ๋ฐ›์•„ ํ˜„์žฌ ์œ„์น˜๋ฅผ ์ถ”์ •ํ•œ๋‹ค. ์นด๋ฉ”๋ผ๋กœ๋ถ€ํ„ฐ ๋กœ๋ด‡์˜ ์œ„์น˜๋ฅผ ์ถ”์ •ํ•˜๊ธฐ ์œ„ํ•ด์„œ ๋น„์ฃผ์–ผ ์˜ค๋„๋ฉ”ํŠธ๋ฆฌ ๊ธฐ๋ฒ•์ด ์ด์šฉ๋˜์—ˆ๋‹ค[9]. ์ƒ์„ฑ๋œ ๊ฒฝ๋กœ์™€ ์ถ”์ •๋œ ํ˜„์žฌ ์œ„์น˜๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ์™ธ์—ฐ์  ๋ชจ๋ธ ์˜ˆ์ธก ์ œ์–ด ์•Œ๊ณ ๋ฆฌ์ฆ˜์€ ์ œ์–ด ์ž…๋ ฅ์„ ๊ณ„์‚ฐํ•˜์—ฌ ์ƒ์„ฑ๋œ ๊ฒฝ๋กœ๋ฅผ ๋”ฐ๋ผ ๋ชฉํ‘œ ์ง€์ ๊นŒ์ง€ ๋กœ๋ด‡์„ ์ด๋™์‹œํ‚จ๋‹ค. ๋กœ๋ด‡์˜ ๊ฐ ์„ผ์„œ, ๋ชจํ„ฐ, ์•Œ๊ณ ๋ฆฌ์ฆ˜๋“ค์€ ๋กœ๋ด‡ ์šด์˜ ์ฒด์ œ (ROS)๊ธฐ๋ฐ˜์œผ๋กœ ๊ฐ ๊ฐ์ฒด๋ณ„๋กœ ๊ตฌํ˜„๋˜์–ด ์—ฐ๊ฒฐ๋œ๋‹ค.

๊ทธ๋ฆผ. 1. ์ž์œจ ์ฃผํ–‰ ์‹œ์Šคํ…œ ๊ตฌ์„ฑ

Fig. 1. The components of autonomous driving system

../../Resources/kiee/KIEE.2019.68.5.678/fig1.png

2.1 ๋‹ค์ค‘ ์„ผ์„œ๊ธฐ๋ฐ˜ ์ƒํƒœ ์ถ”์ •๊ธฐ

์—”์ฝ”๋”์™€ ์นด๋ฉ”๋ผ๋กœ๋ถ€ํ„ฐ ์ธก์ •๋œ ์œ„์น˜ ๊ฐ’์œผ๋กœ๋ถ€ํ„ฐ ๋กœ๋ด‡์˜ ์œ„์น˜ ๋ณ€ํ™”๋ฅผ ์ถ”์ •ํ•˜๊ธฐ ์œ„ํ•˜์—ฌ ํผ์ง€ ๋ชจ๋ธ ๊ธฐ๋ฐ˜์˜ ๋‹ค์ค‘ ์ฃผ๊ธฐ๋ฅผ ๊ฐ€์ง€๋Š” ์ƒํƒœ ์ถ”์ •๊ธฐ๊ฐ€ ์ด์šฉ๋œ๋‹ค. ์ƒํƒœ ์ถ”์ •๊ธฐ๋Š” ์ด๋™ ๋กœ๋ด‡์˜ ์ˆ˜ํ•™์  ๋ชจ๋ธ์„ ๊ธฐ๋ฐ˜์œผ๋กœ ํ˜„์žฌ ์ธก์ •๋œ ๊ฐ’์œผ๋กœ๋ถ€ํ„ฐ ์‹ค์ œ ๋กœ๋ด‡์˜ ์œ„์น˜๋ฅผ ์ถ”์ •ํ•œ๋‹ค. ๋ณธ ์—ฐ๊ตฌ์—์„œ๋Š” ์ด๋™ ๋กœ๋ด‡ ์ถ”์ ์ œ์–ด ๋ชจ๋ธ์ด ์ด์šฉ๋œ๋‹ค[10]. T-S ํผ์ง€ ๋ชจ๋ธ์„ ์ด์šฉํ•˜์—ฌ ๋น„์„ ํ˜• ์ด๋™ ๋กœ๋ด‡ ์ถ”์  ์ œ์–ด ์‹œ์Šคํ…œ์„ ๋ชจ๋ธ๋งํ•˜์˜€๋‹ค. ์ƒํƒœ ์ถ”์ •๊ธฐ์˜ ์ด๋“ ๊ฐ’์„ ์„ค๊ณ„ํ•˜๊ธฐ ์œ„ํ•ด ๋‹ค์ค‘ ์ฃผ๊ธฐ ์ƒ˜ํ”Œ๋ง์„ ๊ณ ๋ คํ•œ ๋ฆฌ์•„ํ”„๋…ธํ”„ ํ•จ์ˆ˜ ๊ธฐ๋ฐ˜์˜ ์„ค๊ณ„ ๋ฐฉ๋ฒ•์ด ์ด์šฉ๋˜์—ˆ๋‹ค[11]. ๊ทธ๋ฆผ. 2์—์„œ์™€ ๊ฐ™์ด ์นด๋ฉ”๋ผ์™€ ์—”์ฝ”๋”๋กœ๋ถ€ํ„ฐ ์ธก์ •๋œ ์„ผ์„œ ๊ฐ’๋“ค์€ ์ƒํƒœ ์ถ”์ •๊ธฐ์— ์˜ํ•ด ํ•˜๋‚˜์˜ ์‹ ํ˜ธ๋กœ ๊ณ„์‚ฐ๋œ๋‹ค.

๊ทธ๋ฆผ. 2. ๋‹ค์ค‘ ์„ผ์„œ๋ฅผ ๊ฐ€์ง€๋Š” ์ƒํƒœ ์ถ”์ •๊ธฐ ๊ตฌ์กฐ

Fig. 2. The structure of state estimotor with multi-sensor

../../Resources/kiee/KIEE.2019.68.5.678/fig2.png

2.2 ์™ธ์—ฐ์  ๋ชจ๋ธ ์˜ˆ์ธก ์ œ์–ด๊ธฐ ์„ค๊ณ„

๋ชจ๋ธ ์˜ˆ์ธก ์ œ์–ด๊ธฐ๋Š” ์‹œ์Šคํ…œ์˜ ์ž…๋ ฅ์˜ ์ œํ•œ์„ ์กฐ๊ฑด์„ ๊ณ ๋ คํ•˜์—ฌ ์ตœ์ ์˜ ์„ฑ๋Šฅ์„ ๋ณด์žฅํ•ด์ฃผ๋Š” ์ œ์–ด๊ธฐ์ด๋‹ค. ๊ทธ๋Ÿฌ๋‚˜ ์‹ค์‹œ๊ฐ„์œผ๋กœ ๋ณต์žกํ•œ ์ตœ์ ํ™” ๋ฌธ์ œ๋ฅผ ํ’€์–ด์„œ ์ž…๋ ฅ์„ ๊ณ„์‚ฐํ•ด์•ผ ํ•˜๋Š” ๋‹จ์ ์ด ์žˆ๋‹ค. ๋”ฐ๋ผ์„œ ์ถฉ๋ถ„ํ•œ ์ปดํ“จํŒ… ์ž์›์ด ์ง€์›๋˜์ง€ ์•Š๋Š” ํ™˜๊ฒฝ์—์„œ๋Š” ์ž…๋ ฅ ๊ณ„์‚ฐ์„ ์œ„ํ•ด ๋งŽ์€ ์‹œ๊ฐ„์ด ์†Œ๋น„๋œ๋‹ค๋Š” ๋‹จ์ ์ด ์žˆ๋‹ค. ์ด๋Ÿฌํ•œ ๋ฌธ์ œ๋ฅผ ํ•ด๊ฒฐํ•˜๊ธฐ ์œ„ํ•˜์—ฌ ์„ค๊ณ„๋œ ์™ธ์—ฐ์  ๋ชจ๋ธ ์˜ˆ์ธก ์ œ์–ด๊ธฐ๋Š” ์˜คํ”„๋ผ์ธ์—์„œ ์ตœ์ ์˜ ์ž…๋ ฅ ๊ฐ’์„ ๋ชจ๋‘ ๊ณ„์‚ฐํ•œ ํ›„, ์‚ฌ์šฉํ•˜๊ธฐ ๋•Œ๋ฌธ์— ๊ณ„์‚ฐ ์ง€์—ฐ ๋ฌธ์ œ๋ฅผ ํ•ด๊ฒฐํ•  ์ˆ˜ ์žˆ๋‹ค.์™ธ์—ฐ์  ๋ชจ๋ธ ์˜ˆ์ธก ๊ธฐ๋ฒ•์—์„œ๋Š” ์˜คํ”„๋ผ์ธ์—์„œ ์ƒํƒœ ๋ณ€์ˆ˜๋“ค์˜ ์˜์—ญ์— ๋”ฐ๋ผ ํ”Œ๋žœํŠธ๋ฅผ ์ œ์–ดํ•จ์— ์žˆ์–ด์„œ ํ•„์š”ํ•œ ์ตœ์ ์˜ ์ž…๋ ฅ ๊ฐ’์„ ๋ชจ๋‘ ๊ณ„์‚ฐํ•œ๋‹ค. ์˜จ๋ผ์ธ์—์„œ๋Š” ๊ณ„์‚ฐ๋œ ์ž…๋ ฅ ๊ฐ’๋“ค์€ ํ•จ์ˆ˜ ํ˜•ํƒœ๋กœ ์ €์žฅ์ด ๋˜์–ด ์žˆ์–ด ํ”Œ๋žœํŠธ์—์„œ ์ƒํƒœ ๋ณ€์ˆ˜๋ฅผ ํ•จ์ˆ˜์— ์ ์šฉํ•˜๋ฉด ๋ณ€์ˆ˜๋“ค์˜ ์˜์—ญ์— ๋งž๋Š” ์ตœ์ ์˜ ์ž…๋ ฅ ๊ฐ’์„ ๋ฐ˜ํ™˜ํ•˜์—ฌ ์ค€๋‹ค. ์™ธ์—ฐ์  ๋ชจ๋ธ ์˜ˆ์ธก ์ œ์–ด๋Š” ํ˜„์žฌ ์ˆœ๊ฐ„์˜ ์ƒํƒœ ๋ณ€์ˆ˜๋ฅผ ์ด์šฉํ•˜์—ฌ ์ž…๋ ฅ ๊ฐ’์„ ๊ณ„์‚ฐํ•˜๋Š” ๊ฒƒ์ด ์•„๋‹ˆ๋ผ, ํ˜„์žฌ ์ƒํƒœ $x_{0}$๊ฐ€ ๊ฐ€์งˆ ์ˆ˜ ์žˆ๋Š” ์ง‘ํ•ฉ $X$ ์— ๋Œ€ํ•˜์—ฌ $u=ax+b$์˜ ์•„ํ•€ ํ˜•ํƒœ๋กœ ์ตœ์  ์ž…๋ ฅ ๊ฐ’์„ ๊ณ„์‚ฐํ•˜๋Š” ๊ฒƒ์ด๋‹ค. ์™ธ์—ฐ์  ๋ชจ๋ธ ์˜ˆ์ธก ์ œ์–ด๊ธฐ๋ฅผ ์„ค๊ณ„ํ•˜๊ธฐ ์œ„ํ•ด์„œ M. Kvasnica and P. Grieder and M. Baoti์— ์˜ํ•ด ์ œ๊ณต๋œ Multi-Parametric Toolbox (MPT)๋ฅผ ์ด์šฉํ•˜์˜€๋‹ค[8]. MPT๋Š” ๊ธฐ๋ณธ์ ์ธ ์ผ์ฐจ ์ด์ฐจ ๊ณ„ํš๋ฒ•์„ ํ•ด๊ฒฐํ•˜๋Š” ํ•จ์ˆ˜๋“ค์„ ์ œ๊ณตํ•ด์ค„ ๋ฟ๋งŒ ์•„๋‹ˆ๋ผ, ์™ธ์—ฐ์  ๋ชจ๋ธ ์˜ˆ์ธก ์ œ์–ด๊ธฐ ์„ค๊ณ„ ๋‚ด์žฅ ํ•จ์ˆ˜์™€ ๋‹ค๋ฅธ ๊ฐœ๋ฐœ ์–ธ์–ด๋กœ ๋ณ€ํ™˜์‹œ์ผœ์ฃผ๋Š” ๋‚ด์žฅ ํ•จ์ˆ˜๋“ค์„ ์ œ๊ณตํ•˜๊ณ  ์žˆ๋‹ค.

ํ‘œ 1์€ MPT๋ฅผ ์ด์šฉํ•˜์—ฌ ๋งคํŠธ๋žฉ์—์„œ ์™ธ์—ฐ์  ๋ชจ๋ธ ์˜ˆ์ธก ์ œ์–ด๊ธฐ๋ฅผ ์„ค๊ณ„ํ•˜์—ฌ Python ์–ธ์–ด๋กœ ๋ณ€ํ™˜์‹œ์ผœ์ฃผ๋Š” ๋‹จ๊ณ„๋ฅผ ์„ค๋ช…ํ•˜์˜€๋‹ค. 1 ๋‹จ๊ณ„์—์„œ๋Š” ์ด๋™ ๋กœ๋ด‡ ๋ชจ๋ธ์˜ ์ด์‚ฐ์  ์ƒํƒœ ๋ณ€์ˆ˜ ๋ฐฉ์ •์‹์˜ A, B ํ–‰๋ ฌ์„ ์„ค์ •ํ•ด์ฃผ๋ฉฐ, 2 ๋‹จ๊ณ„์—์„œ๋Š” ์ƒํƒœ๋ณ€์ˆ˜๋“ค๊ณผ ์ž…๋ ฅ์˜ ํฌ๊ธฐ ์ œํ•œ์„ ๊ณ ๋ คํ•œ๋‹ค. 3 ๋‹จ๊ณ„์—์„œ๋Š” ์˜ˆ์ธก ์ œ์–ด๊ธฐ์˜ ์ž…๋ ฅ ๋ฏธ๋ž˜ ์˜ˆ์ธก ๊ตฌ๊ฐ„์„ ์„ค์ •ํ•œ๋‹ค. 4 ๋‹จ๊ณ„์—์„œ๋Š” ๋ชจ๋ธ ์˜ˆ์ธก ์ œ์–ด๊ธฐ๋ฅผ ๋‚ด์žฅ ํ•จ์ˆ˜๋กœ ๋จผ์ € ์ƒ์„ฑ ํ•œ ๋‹ค์Œ ctrl.toExplicit()๋ฅผ ์ด์šฉํ•˜์—ฌ ์™ธ์—ฐ์  ๋ชจ๋ธ ์˜ˆ์ธก ์ œ์–ด๊ธฐ๋ฅผ ์ƒ์„ฑํ•œ๋‹ค. explicit_ctrl.partition.plot()๋ฅผ ์ด์šฉํ•˜๋ฉด ์ƒํƒœ ๋ณ€์ˆ˜๋“ค์˜ ์˜์—ญ์— ๋”ฐ๋ฅธ ์ตœ์ ์˜ ์ž…๋ ฅ ๊ฐ’์„ ๊ทธ๋ž˜ํ”„๋กœ ๋‚˜ํƒ€๋‚ธ๋‹ค. ๊ทธ๋ฆผ. 3์€ 3์ฐจ ์ƒํƒœ ๋ณ€์ˆ˜ ๋ฐฉ์ •์‹์˜ ์ด๋™ ๋กœ๋ด‡ ๋ชจ๋ธ์„ ์ด์šฉํ•˜์—ฌ ์ž…๋ ฅ ์ œํ•œ ์กฐ๊ฑด์„ ๊ณ ๋ คํ•˜์—ฌ ๊ตฌํ•œ ์ตœ์ ์˜ ์ž…๋ ฅ ๊ฐ’ ๊ทธ๋ž˜ํ”„์ด๋‹ค. ๊ฐ ์˜์—ญ์— ๋”ฐ๋ฅธ ์ตœ์ ์˜ ์ž…๋ ฅ ๊ฐ’์€ ๋‹ค๋ฅธ ์ƒ‰์œผ๋กœ ํ‘œ์‹œ๋˜์–ด ์žˆ์œผ๋ฉฐ, ๊ฐ ์˜์—ญ ๋ณ„๋กœ ์ž…๋ ฅ ๊ฐ’์„ ๊ตฌํ•˜๊ธฐ ์œ„ํ•ด์„œ๋Š” ์•„๋ž˜์˜ ์‹์—์„œ

ํ‘œ 1. MPT๋ฅผ ์ด์šฉํ•œ ์™ธ์—ฐ์  ๋ชจ๋ธ ์˜ˆ์ธก ์ œ์–ด๊ธฐ ์„ค๊ณ„ ๋‹จ๊ณ„ ๋ฐ ํŠน์ง•

Table 1. Design procedure and characteristics of explicit model predictive contoller using MPT

๋‹จ๊ณ„

MPT ๋‚ด์žฅ ํ•จ์ˆ˜

1. ๋ชจ๋ธ ์ƒ์„ฑ

model = LTISystem('A', Ad, 'B', Bd)

2. ์ œํ•œ ์กฐ๊ฑด ๊ณ ๋ ค

model.x.min = [(1); (2); (3)]

model.x.max = [(1); (2); (3)]

model.u.min = [(1); (2)]

model.u.max = [(1); (2)]

3. ์˜ˆ์ธก ๊ตฌ๊ฐ„ ์„ค์ •

horizon=value

4. ์ œ์–ด๊ธฐ ์ƒ์„ฑ ๋ฐ ์ƒํƒœ ๋ณ€์ˆ˜ ์˜์—ญ ํ™•์ธ

ctrl = MPCController(model, horizon);

explicit_ctrl = ctrl.toExplicit()

explicit_ctrl.partition.plot()

5. Python ์ถ”์ถœ

opj = explicit_ctrl.optimizer;

opj.toPython('filename','primal','obj')

๊ทธ๋ฆผ. 3. ์ƒํƒœ ๋ณ€์ˆ˜๋“ค์˜ ๋‹ค๊ฐํ˜• ์˜์—ญ์— ๋”ฐ๋ฅธ ์ตœ์ ์˜ ์ž…๋ ฅ ๊ฐ’ ๊ทธ๋ž˜ํ”„

Fig. 3. Optimal input value graph according to polygonal area of state variables

../../Resources/kiee/KIEE.2019.68.5.678/fig3.png

$\begin{bmatrix}x_{1} & 1\\ x_{2} & 1\end{bmatrix}\begin{bmatrix}a \\ b\end{bmatrix}=\begin{bmatrix}u_{1}\\ u_{2}\end{bmatrix}$

์—ญํ–‰๋ ฌ์„ ์ด์šฉํ•˜์—ฌ a์™€ b์˜ ๋ณ€์ˆ˜๋“ค์„ ์ฐพ์•„์•ผ ํ•œ๋‹ค. ์ด๋ฅผ ๋‚ด์žฅ ํ•จ์ˆ˜ ํ˜•ํƒœ๋กœ ๊ตฌํ˜„ ํ•œ ๊ฒƒ์ด ctrl.evaluate(x0) ํ•จ์ˆ˜์ด๋‹ค. ์ƒํƒœ ๋ณ€์ˆ˜์˜ ํ˜„์žฌ ๊ฐ’๋งŒ ์„ค์ •ํ•ด์ฃผ๋ฉด ์ตœ์ ์˜ ์ž…๋ ฅ ๊ฐ’์„ ์ž๋™ ๊ณ„์‚ฐํ•˜์—ฌ ๊ฒฐ๊ณผ ๊ฐ’์œผ๋กœ ๋‚˜ํƒ€๋‚ธ๋‹ค. ๋งˆ์ง€๋ง‰ 5 ๋‹จ๊ณ„์—์„œ๋Š” MATLAB์—์„œ ์ƒ์„ฑ๋œ ์ œ์–ด๊ธฐ๋ฅผ Python ์–ธ์–ด๋กœ ๋ณ€ํ™˜ ์‹œ์ผœ์ฃผ๋Š” ๊ณผ์ •์ธ opj.toPython ('filename','primal','obj')์˜ ๋‚ด์žฅ ํ•จ์ˆ˜๋ฅผ ์ด์šฉํ•˜์˜€๋‹ค.

2.3 ์นด๋ฉ”๋ผ ์˜์ƒ ์‹ ํ˜ธ ์ฒ˜๋ฆฌ

์ œ์•ˆ๋œ ์ž์œจ ์ฃผํ–‰ ์‹œ์Šคํ…œ์—์„œ ์˜์ƒ ๋ฐ์ดํ„ฐ๋Š” ๋กœ๋ด‡์˜ ์œ„์น˜ ๋ณ€ํ™” ์ธก์ •๊ณผ ํ‘œ์ง€ํŒ ๊ฒ€์ถœ 2๊ฐ€์ง€ ๊ธฐ๋Šฅ์„ ์œ„ํ•ด ์‚ฌ์šฉ๋œ๋‹ค. ์˜์ƒ์„ ํ†ตํ•œ ๋กœ๋ด‡์˜ ์œ„์น˜ ๋ณ€ํ™” ์ธก์ •์„ ์œ„ํ•œ ๋ฐฉ๋ฒ•์œผ๋กœ Zed ์นด๋ฉ”๋ผ์— ๊ตฌํ˜„๋˜์–ด์ง„ ๋น„์ฃผ์–ผ ์˜ค๋„๋ฉ”ํŠธ๋ฆฌ ๊ธฐ๋ฒ•์„ ์‚ฌ์šฉํ•˜์˜€๋‹ค[9]. ๋น„์ฃผ์–ผ ์˜ค๋„๋ฉ”ํŠธ๋ฆฌ๋Š” ์ด๋ฏธ์ง€๋“ค ๊ฐ„์˜ ํŠน์ง•์  ๋ณ€ํ™”๋ฅผ ๊ณ„์‚ฐํ•˜์—ฌ ์นด๋ฉ”๋ผ์˜ ์ด๋™๋Ÿ‰์„ ์ธก์ •ํ•˜๋Š” ๊ธฐ์ˆ ์ด๋‹ค. ์‚ฌ์šฉ๋œ ๋น„์ฃผ์–ผ ์˜ค๋„๋ฉ”ํŠธ๋ฆฌ์˜ ์•Œ๊ณ ๋ฆฌ์ฆ˜์€ ํŠน์ง•์  ์ถ”์ถœ, ์•„์›ƒ๋ผ์ด์–ด ์ œ๊ฑฐ, ํŠน์ง•์  ๋งค์นญ, ์นด๋ฉ”๋ผ ํŒŒ๋ผ๋ฏธํ„ฐ๋ฅผ ์ด์šฉํ•œ ์‹ค์ œ ์ด๋™๊ฑฐ๋ฆฌ ๊ณ„์‚ฐ์œผ๋กœ ๊ตฌ์„ฑ๋˜์–ด ์žˆ๋‹ค. ํ‘œ์ง€ํŒ ๊ฒ€์ถœ ์•Œ๊ณ ๋ฆฌ์ฆ˜์œผ๋กœ opencv์˜ TM_SQDIFF_NORMED ๋ฐฉ๋ฒ•์„ ๊ธฐ๋ฐ˜์œผ๋กœ ๋‹ค์ค‘ ํ…œํ”Œ๋ฆฟ ๋งค์นญ ๋ฐฉ๋ฒ•์ด ์ œ์•ˆ๋˜์—ˆ๋‹ค. ํ…œํ”Œ๋ฆฟ ๋ฐ์ดํ„ฐ ํš๋“ ๊ณผ์ •์—์„œ๋Š” ํ™˜๊ฒฝ์˜ ๋ณ€ํ™”์—๋„ ๊ฐ•์ธํ•œ ํ‘œ์ง€ํŒ ๊ฒ€์ถœ ์„ฑ๋Šฅ์„ ๋ณด์žฅํ•˜๊ธฐ ์œ„ํ•ด ๋‹ค์–‘ํ•œ ํ™˜๊ฒฝ์—์„œ ํš๋“ํ•œ ํ‘œ์ง€ํŒ ์˜์ƒ์—์„œ ํŠน์ง•์ด ๋˜๋Š” ๋ถ€๋ถ„๋“ค๋กœ ํ…œํ”Œ๋ฆฟ์ด ๋งŒ๋“ค์–ด์ง„๋‹ค. ๊ฒ€์ถœ ๊ณผ์ •์—์„œ๋Š” ํ˜„์žฌ ํš๋“ํ•œ ์ด๋ฏธ์ง€์™€ ํ…œํ”Œ๋ฆฟ ๊ฐ„์˜ ์ฐจ์ด๋„๊ฐ€ ๊ณ„์‚ฐ๋œ๋‹ค. ํ•„ํ„ฐ๋ง ๊ณผ์ •์—์„œ๋Š” ํ…œํ”Œ๋ฆฟ๊ณผ ๋“ค์–ด์˜จ ์ด๋ฏธ์ง€์˜ ์ฐจ์ด๋„ ๊ฐ’์ด ์ผ์ • ๊ฐ’ ์ดํ•˜์ด๋ฉด ๊ฒ€์ถœ ํ›„๋ณด ์˜์—ญ์œผ๋กœ ์ง€์ •ํ•œ๋‹ค. ๊ฐ ํ…œํ”Œ๋ฆฟ์—์„œ ์ง€์ •๋œ ํ›„๋ณด ์˜์—ญ๋“ค ๊ฐ„์˜ ๊ฑฐ๋ฆฌ ์ฐจ์ด๊ฐ€ ๋ฏธ๋ฆฌ ์ •ํ•ด์ง„ ์ž„๊ณ„๊ฐ’ ๋ณด๋‹ค ์ž‘์€ ๊ฒฝ์šฐ์—๋Š” ํ‘œ์ง€ํŒ์ด ๊ฒ€์ถœ๋˜์—ˆ๋‹ค๊ณ  ํŒ๋‹จํ•˜๊ณ  ์ž„๊ณ„๊ฐ’ ๋ณด๋‹ค ํฐ ๊ฒฝ์šฐ์—๋Š” ํ‘œ์ง€ํŒ์ด ์—†๋Š” ๊ฒฝ์šฐ๋กœ ํŒ๋‹จํ•œ๋‹ค. ์ œ์•ˆ๋œ ๋ฐฉ๋ฒ•์€ ์ž‘์€ ํ…œํ”Œ๋ฆฟ์„ ์ด์šฉํ•˜์—ฌ ๊ณ„์‚ฐ ์‹œ๊ฐ„์„ ์ค„์ผ ์ˆ˜ ์žˆ๊ณ , ์ถ”๊ฐ€์ ์ธ ์ •๋ณด์™€ ๋ชฉํ‘œ ๋ฌผ์ฒด์˜ ํŠน์ง•์„ ๋‚˜ํƒ€๋‚ด๋Š” ํ…œํ”Œ๋ฆฟ์„ ์‚ฌ์šฉํ•จ์œผ๋กœ ์ ์€ ์–‘์˜ ๋ฐ์ดํ„ฐ๋ฅผ ํ†ตํ•ด ๊ธฐ์กด์˜ ํ…œํ”Œ๋ฆฟ ๋งค์นญ ๋ฐฉ๋ฒ•๋ณด๋‹ค ๊ฒ€์ถœ ์„ฑ๋Šฅ์„ ๋†’์ผ ์ˆ˜ ์žˆ๋‹ค.

๊ทธ๋ฆผ. 4. ํ‘œ์ง€ํŒ ๊ฒ€์ถœ ์•Œ๊ณ ๋ฆฌ์ฆ˜

Fig. 4. Sign detection algorithm

../../Resources/kiee/KIEE.2019.68.5.678/fig4.png

2.4 ๋ผ์ด๋‹ค ์‹ ํ˜ธ ์ฒ˜๋ฆฌ

์ด๋™๋กœ๋ด‡์˜ ์ฃผ๋ณ€์˜ ์žฅ์• ๋ฌผ์„ ๊ฒ€์ถœํ•˜๊ณ  ๊ฒฝ๋กœ๋ฅผ ์ƒ์„ฑํ•˜๊ธฐ ์œ„ํ•˜์—ฌ ๋ผ์ด๋” ์„ผ์„œ (Velodyne LiDAR)๊ฐ€ ์ด์šฉ๋œ๋‹ค. 16 ์ฑ„๋„์˜ ๋ฐ์ดํ„ฐ๋ฅผ 2์ฐจ์›์˜ ์žฅ์• ๋ฌผ ์œ„์น˜ ์ •๋ณด ๋ฐ์ดํ„ฐ๋กœ ๋ณ€ํ™˜ํ•˜๊ธฐ ์œ„ํ•˜์—ฌ ๊ทธ๋ฆผ. 5์™€ ๊ฐ™์€ ๊ณผ์ •์„ ๊ฑฐ์นœ๋‹ค.

๊ทธ๋ฆผ. 5. ๋ผ์ด๋‹ค ๋ฐ์ดํ„ฐ ์ฒ˜๋ฆฌ ๊ณผ์ •

Fig. 5. LiDAR data processing procedure

../../Resources/kiee/KIEE.2019.68.5.678/fig5.png

์ฒซ ๋ฒˆ์งธ ๊ณผ์ •์œผ๋กœ ๋ผ์ด๋”๋กœ๋ถ€ํ„ฐ ๊ฐ ์ฑ„๋„์— ๋Œ€ํ•˜์—ฌ ์ฃผ๋ณ€์˜ ๋ฌผ์ฒด์™€์˜ ๊ฑฐ๋ฆฌ์ •๋ณด๋ฅผ ์ง๊ตํ‰๋ฉด์œผ๋กœ ํ‘œํ˜„๋œ ๋ฐ์ดํ„ฐ๋ฅผ ํš๋“ํ•œ๋‹ค. ๊ทธ ๋‹ค์Œ ๊ฐ ํ•„์š”ํ•œ ์ฑ„๋„๋งŒ์˜ ๋ฐ์ดํ„ฐ๋“ค์„ ์„ ํƒํ•˜๊ณ  ๊ฐ ์ฑ„๋„์˜ ๋™์ผํ•œ ์œ„์น˜์— ์žˆ๋Š” ๊ฐ’๋“ค์„ ๋ˆ„์ ํ•˜์—ฌ ํ•˜๋‚˜์˜ 2์ฐจ์› ํ‰๋ฉด ๋ฐ์ดํ„ฐ๋กœ ๋ณ€ํ™˜ํ•œ๋‹ค. ์ด๋•Œ ์ผ์ • ๊ฐ’ ์ดํ•˜์˜ ๊ฐ’๋“ค์€ ๋…ธ์ด์ฆˆ๋กœ ๊ฐ„์ฃผํ•˜์—ฌ ์ œ๊ฑฐํ•œ๋‹ค. ๋ฌผ์ฒด๋ฅผ ๊ฒ€์ถœํ•˜๊ธฐ ์œ„ํ•œ ๊ณผ์ •์œผ๋กœ ๋ ˆ์ด๋ธ”๋ง ๊ณผ์ •์„ ๊ฑฐ์ณ ์ผ์ • ๋ฉ์–ด๋ฆฌ ์ดํ•˜์˜ ๊ฐ’๋“ค์„ ์ œ๊ฑฐํ•œ๋‹ค. ๋กœ๋ด‡์ด ์ด๋™ํ•ด์•ผํ•  ์œ„์น˜๋ฅผ ์ •ํ•˜๊ธฐ ์œ„ํ•˜์—ฌ ์žฅ์• ๋ฌผ๊ฐ„์˜ ์ค‘์ ์„ ๊ณ„์‚ฐํ•˜์—ฌ ๋ชฉํ‘œ ์ง€์ ์„ ์ง€์ •ํ•œ๋‹ค.

3. ์ž์œจ ์ด๋™๋กœ๋ด‡ ์ถ”์ ์ œ์–ด ์‹œ์Šคํ…œ ๊ตฌํ˜„

3.1 ๋กœ๋ด‡ ์šด์˜์ฒด์ œ ๊ธฐ๋ฐ˜์˜ ์ž„๋ฒ ๋””๋“œ ์‹œ์Šคํ…œ ๊ตฌํ˜„

์ด๋™ ๋กœ๋ด‡๊ณผ ์นด๋ฉ”๋ผ ๋ฐ ๋ผ์ด๋‹ค๋ฅผ ์ œ์–ดํ•˜๊ณ  ์‹ ํ˜ธ ์ฒ˜๋ฆฌ ๋ฐ ์ œ์–ด ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ๊ตฌํ˜„ํ•˜๊ณ  ๊ด€๋ฆฌํ•˜๊ธฐ ์œ„ํ•˜์—ฌ ๋กœ๋ด‡ ์šด์˜ ์ฒด์ œ ๊ธฐ๋ฐ˜์˜ ์‹œ์Šคํ…œ์„ ๊ตฌํ˜„ํ•˜์˜€๋‹ค. ๋กœ๋ด‡ ์šด์˜์ฒด์ œ(ROS)๋Š” ๋กœ๋ด‡ ์‘์šฉ ํ”„๋กœ๊ทธ๋žจ์„ ๊ฐœ๋ฐœํ•  ๋•Œ ํ•„์š”ํ•œ ํ•˜๋“œ์›จ์–ด ์ถ”์ƒํ™”, ํ•˜์œ„ ๋””๋ฐ”์ด์Šค ์ œ์–ด, ์ผ๋ฐ˜์ ์œผ๋กœ ์‚ฌ์šฉ๋˜๋Š” ๊ธฐ๋Šฅ์˜ ๊ตฌํ˜„, ํ”„๋กœ์„ธ์Šค ๊ฐ„์˜ ๋ฉ”์‹œ์ง€ ํŒจ์‹ฑ, ํŒจํ‚ค์ง€ ๊ด€๋ฆฌ, ๊ฐœ๋ฐœํ™˜๊ฒฝ์— ํ•„์š”ํ•œ ๋ผ์ด๋ธŒ๋Ÿฌ๋ฆฌ์™€ ๋‹ค์–‘ํ•œ ๊ฐœ๋ฐœ ๋ฐ ๋””๋ฒ„๊น… ๋„๊ตฌ๋ฅผ ์ œ๊ณตํ•œ๋‹ค. ๊ฐ ํ•˜๋“œ์›จ์–ด ์š”์†Œ๋“ค๊ณผ ๊ธฐ๋Šฅ๋“ค์€ ๋…ธ๋“œ ๋‹จ์œ„๋กœ ๊ตฌํ˜„๋˜๊ณ  ํ† ํ”ฝ์„ ์ด์šฉํ•˜์—ฌ ๋ฐ์ดํ„ฐ๋ฅผ ์ฃผ๊ณ ๋ฐ›๋Š”๋‹ค. ์ œ์•ˆ๋œ ์‹œ์Šคํ…œ์€ ๋…ธ๋“œ๋“ค์„ ๊ด€๋ฆฌํ•˜๋Š” ๋…ธ๋“œ, ๋กœ๋ด‡ ๋ชจํ„ฐ ๋ฐ ์—”์ฝ”๋” ์ œ์–ด ๋…ธ๋“œ, ๋ผ์ด๋‹ค ์„ผ์„œ๋ฅผ ๊ด€๋ฆฌํ•˜๋Š” ๋ผ์ด๋‹ค ๋…ธ๋“œ, ์˜์ƒ ๋ฐ์ดํ„ฐ๋ฅผ ํš๋“ํ•˜๊ณ  ๋น„์ฃผ์–ผ ์˜ค๋„๋ฉ”ํŠธ๋ฆฌ๋ฅผ ๊ณ„์‚ฐํ•˜๋Š” ์นด๋ฉ”๋ผ ๋…ธ๋“œ, ํ‘œ์ง€ํŒ ๊ฒ€์ถœ ๋…ธ๋“œ, ์ƒํƒœ ์ถ”์ •๊ธฐ ๋ฐ ์ œ์–ด ์ž…๋ ฅ ๊ณ„์‚ฐ ๋…ธ๋“œ, ๋ผ์ด๋‹ค ๋ฐ์ดํ„ฐ ์ฒ˜๋ฆฌ ๋…ธ๋“œ๋กœ ๊ตฌ์„ฑ๋˜์–ด ์žˆ๋‹ค. ๊ทธ๋ฆผ. 6์€ ์ž„๋ฒ ๋””๋“œ ๋ณด๋“œ (NVIDIA Jetson TX2)์— ๊ตฌํ˜„๋œ ๋…ธ๋“œ๋“ค์˜ ์—ฐ๊ฒฐ ๊ตฌ์„ฑ๊ณผ ํ† ํ”ฝ์„ ํ†ตํ•ด ์†ก์ˆ˜์‹ ๋˜๋Š” ๋ฐ์ดํ„ฐ๋“ค์„ ๋‚˜ํƒ€๋‚ธ๋‹ค.

๊ทธ๋ฆผ. 6. ๊ตฌํ˜„๋œ ๋กœ๋ด‡ ์šด์˜ ์ฒด์ œ์˜ ๋…ธ๋“œ ์—ฐ๊ฒฐ ๊ตฌ์„ฑ

Fig. 6. Node connection structure of impremeted robot operating system (ROS)

../../Resources/kiee/KIEE.2019.68.5.678/fig6.png

3.2 ์ž์œจ ์ด๋™ ๋กœ๋ด‡ ์ถ”์  ์ œ์–ด ์‹œ์Šคํ…œ

๋‹ค์ค‘ ์„ผ์„œ๋ฅผ ์ด์šฉํ•˜๋Š” ์ƒํƒœ ์ถ”์ •๊ธฐ, ์™ธ์—ฐ์  ๋ชจ๋ธ ์˜ˆ์ธก ๊ธฐ๊ธฐ๋ฐ˜์˜ ์ถ”์  ์ œ์–ด ์‹œ์Šคํ…œ์„ ๊ตฌํ˜„ํ•˜๊ธฐ ์œ„ํ•ด ์ž„๋ฒ ๋””๋“œ ๋ณด๋“œ (NVIDIA Jetson TX2)์— ์•Œ๊ณ ๋ฆฌ์ฆ˜์ด ๊ตฌํ˜„๋˜์—ˆ๋‹ค. ๊ทธ๋ฆผ. 7์€ ์ž์œจ ์ฃผํ–‰ ์‹œ์Šคํ…œ์˜ ๊ตฌ์กฐ๋ฅผ ๋‚˜ํƒ€๋‚ธ๋‹ค. ์ฃผํ–‰์ด ์‹œ์ž‘๋˜๋ฉด ๋กœ๋ด‡์€ ๋ผ์ด๋”๋กœ๋ถ€ํ„ฐ ๋ฐ์ดํ„ฐ๋ฅผ ๋ฐ›์•„ ๋กœ๋ด‡์ด ์ฃผํ–‰ํ•  ๊ฒฝ๋กœ๋ฅผ ๊ณ„ํšํ•œ๋‹ค. ๋ผ์ด๋” ๋ฐ์ดํ„ฐ์ฒ˜๋ฆฌ ํ›„ ์ƒ์„ฑ๋œ ๋ชฉํ‘œ์ง€์ ์˜ ์ขŒํ‘œ๊ฐ€ ์ฃผ์–ด์ง€๋ฉด ๋ชจ๋ฐ”์ผ ๋กœ๋ด‡์˜ ํ‚ค๋„ค๋งˆํ‹ฑ ๋ชจ๋ธ์„ ๊ธฐ๋ฐ˜์œผ๋กœ ๋กœ๋ด‡์˜ ๊ถค์ ์„ ๊ณ„์‚ฐํ•œ๋‹ค. ๊ถค์  ์ƒ์„ฑ์„ ์œ„ํ•œ ๋กœ๋ด‡์˜ ์ž…๋ ฅ์€ ๋‹ค์Œ๊ณผ ๊ฐ™์ด ์ฃผ์–ด์ง„๋‹ค.

๊ทธ๋ฆผ. 7. ์ด๋™ ๋กœ๋ด‡์˜ ์ž์œจ์ฃผํ–‰ ์‹œ์Šคํ…œ ์†Œํ”„ํŠธ์›จ์–ด ๊ตฌ์กฐ

Fig. 7. Software architecture for an autonomous driving system

../../Resources/kiee/KIEE.2019.68.5.678/fig7.png

$v=K_{1}\rho\cos\Phi$

$w = -K_{1}\sin\Phi\cos\Phi -K_{2}\Phi$

$rho=\sqrt{(x_{t}-x_{i})^{2}+(y_{t}-y_{i})^{2}}$

$\Phi =\theta_{i}-arc(\dfrac{y_{t}-y_{i}}{x_{t}-x_{i}})$

$K_{1} > 0, K_{2} > 0$

$x_{t}$, $x_{i}$๋Š” ๋กœ๋ด‡์˜ ๋ชฉํ‘œ ์ง€์ ์˜ $x$์ขŒํ‘œ์™€ ๋กœ๋ด‡์˜ ํ˜„์žฌ ์œ„์น˜๋ฅผ ๋‚˜ํƒ€๋‚ด๊ณ  $y_{t}$, $y_{i}$๋Š” ๋กœ๋ด‡์˜ ๋ชฉํ‘œ ์ง€์ ์˜ $y$์ขŒํ‘œ์™€ ๋กœ๋ด‡์˜ ํ˜„์žฌ ์œ„์น˜๋ฅผ ๋‚˜ํƒ€๋‚ธ๋‹ค. $\theta_{i}$๋Š” ๋กœ๋ด‡์ด ๋ฐ”๋ผ๋ณด๋Š” ๊ฐ๋„๋ฅผ ๋‚˜ํƒ€๋‚ธ๋‹ค. ์„ ์†๋„ $v$์™€ ๊ฐ์†๋„ $w$๋ฅผ ๋‹ค์Œ์˜ ๋ชจ๋ฐ”์ผ ๋กœ๋ด‡์˜ ํ‚ค๋„ค๋งˆํ‹ฑ ๋ชจ๋ธ์— ์ž…๋ ฅ์œผ๋กœ ์ฃผ์–ด ๋กœ๋ด‡์ด ์ด๋™ํ•ด์•ผํ•  ๊ฒฝ๋กœ์™€ ์†๋„๋ฅผ ๊ณ„์‚ฐํ•œ๋‹ค.

$\dot x =v\cos\theta ,\:\dot y =v\sin\theta ,\:\dot\theta =w.$

๊ฒฝ๋กœ๋Š” ํ‘œ์ง€ํŒ์ด ์žˆ๋Š” ์œ„์น˜๊นŒ์ง€ ์žฅ์• ๋ฌผ์„ ํ”ผํ•ด๊ฐ€๋ฉฐ ์ง์ง„ํ•  ์ˆ˜ ์žˆ๋Š” ๊ฒฝ๋กœ๊ฐ€ ์ฃผ์–ด์ง„๋‹ค. ํ‘œ์ง€ํŒ์ด ์žˆ๋Š” ๊ฒฝ์šฐ ํ‘œ์ง€ํŒ์— ์ œ์‹œํ•˜๋Š” ๋ฐฉํ–ฅ์œผ๋กœ ํšŒ์ „ํ•œ๋‹ค. ๊ณ„ํš๋œ ๊ฒฝ๋กœ์—๋Š” ๋กœ๋ด‡์ด ์ด๋™ํ•ด์•ผํ•  ์œ„์น˜์™€ ์†๋„ ๊ฐ’์ด ๊ณ„์‚ฐ๋˜์–ด์ ธ์žˆ๋‹ค. ๊ฒฝ๋กœ๊ฐ€ ๊ณ„ํš๋˜๋ฉด ๋กœ๋ด‡์˜ ํ˜„์žฌ์œ„์น˜ ๊ฐ’์ด ์นด๋ฉ”๋ผ์™€ ์—”์ฝ”๋”๋กœ๋ถ€ํ„ฐ ์ธก์ •๋˜๊ณ  ์ธก์ •๋œ ๊ฐ’๋“ค์€ ์ƒํƒœ ์ถ”์ •๊ธฐ์˜ ์ž…๋ ฅ์ด ๋˜์–ด ๋กœ๋ด‡์˜ ํ˜„์žฌ ์œ„์น˜๋ฅผ ์ถ”์ •ํ•œ๋‹ค. ์™ธ์—ฐ์  ๋ชจ๋ธ ์˜ˆ์ธก์ œ์–ด๊ธฐ๋Š” ๊ณ„ํš๋œ ๊ฒฝ๋กœ์™€ ๋กœ๋ด‡์˜ ํ˜„์žฌ ์œ„์น˜ ๊ฐ’์„ ์ด์šฉํ•˜์—ฌ ๋กœ๋ด‡์˜ ์ œ์–ด ์ž…๋ ฅ์„ ๊ณ„์‚ฐํ•œ๋‹ค.

4. ์‹คํ—˜ ๊ฒฐ๊ณผ

4.1 ๊ฐ€์ƒ ํ™˜๊ฒฝ ์‹คํ—˜

์ œ์•ˆ๋œ ์‹œ์Šคํ…œ์˜ ์„ฑ๋Šฅ์„ ํ‰๊ฐ€ํ•˜๊ธฐ ์œ„ํ•ด์„œ ๊ฐ€์ œ๋ณด ๊ฐ€์ƒํ™˜๊ฒฝ์—์„œ ์ œ์–ด ์‹œ์Šคํ…œ์˜ ์„ฑ๋Šฅ์„ ํ‰๊ฐ€ํ•˜์˜€๋‹ค. ์„ผ์„œ์˜ ์ธก์ • ๋…ธ์ด์ฆˆ๊ฐ€ ์žˆ๋Š” ์ƒํ™ฉ์„ ๊ณ ๋ คํ•˜๊ธฐ ์œ„ํ•˜์—ฌ ์—”์ฝ”๋”์˜ ๊ฒฝ์šฐ ์œ„์น˜ ์ธก์ •์‹œ -0.05m์—์„œ 0.05m ์‚ฌ์ด์˜ ์ธก์ • ์˜ค์ฐจ๋ฅผ ๋ˆ„์ ๋˜๋„๋ก ๋ชจ๋ธ๋ง๋˜์—ˆ๊ณ  ์นด๋ฉ”๋ผ์˜ ๊ฒฝ์šฐ๋Š” ์œ„์น˜ ์ธก์ •์‹œ โ€“0.005m์—์„œ 0.005m ์‚ฌ์ด ์ธก์ • ์˜ค์ฐจ๊ฐ€ ๋ˆ„์ ๋œ๋‹ค. ์—”์ฝ”๋”๋Š” 0.05s์˜ ์ƒ˜ํ”Œ๋ง ์ฃผ๊ธฐ๋ฅผ ๊ฐ€์ง€๊ณ  ์นด๋ฉ”๋ผ๋Š” 0.5s์˜ ์ƒ˜ํ”Œ๋ง ์ฃผ๊ธฐ๋ฅผ ๊ฐ€์ง„๋‹ค. ๊ทธ๋ฆผ. 8์€ ์นด๋ฉ”๋ผ์™€ ์—”์ฝ”๋”๋กœ๋ถ€ํ„ฐ ์ธก์ •๋œ ๋กœ๋ด‡์˜ ์œ„์น˜์˜ ์ธก์ • ์˜ค์ฐจ($\sqrt{(์‹ค์ œ์œ„์น˜-์ธก์ •์œ„์น˜)^{2}}$)์™€ ์ƒํƒœ ์ถ”์ •๊ธฐ๋กœ๋ถ€ํ„ฐ ์ถ”์ •๋œ ๋กœ๋ด‡์˜ ์œ„์น˜ ์ถ”์ • ์˜ค์ฐจ($\sqrt{(์‹ค์ œ์œ„์น˜-์ถ”์ •์œ„์น˜)^{2}}$)๋ฅผ ๋ณด์—ฌ์ค€๋‹ค. ๊ทธ๋ฆผ. 9๋Š” ์นด๋ฉ”๋ผ์™€ ์—”์ฝ”๋” ๊ฐ๊ฐ์„ ์‚ฌ์šฉํ•œ ๊ฒฝ์šฐ์™€ ์นด๋ฉ”๋ผ์™€ ์—”์ฝ”๋”๋ฅผ ํ•จ๊ป˜ ์‚ฌ์šฉํ•œ ๊ฒฝ์šฐ์˜ ๋กœ๋ด‡ ์ถ”์ ์ œ์–ด ์˜ค์ฐจ($\sqrt{(๋ชฉํ‘œ์œ„์น˜-ํ˜„์žฌ์œ„์น˜)^{2}}$)๋ฅผ ๋ณด์—ฌ์ค€๋‹ค. ์—”์ฝ”๋”๋งŒ์„ ์‚ฌ์šฉํ•œ ๊ฒฝ์šฐ๋Š” ์ธก์ • ์˜ค์ฐจ๊ฐ€ ํฐ ์ด์œ ๋กœ ๊ฒฝ๋กœ ์ถ”์  ์˜ค์ฐจ๊ฐ€ ๋ฐœ์‚ฐํ•œ๋‹ค. ์นด๋ฉ”๋ผ๋งŒ์„ ์‚ฌ์šฉํ•œ ๊ฒฝ์šฐ์—๋Š” ๋Š๋ฆฐ ์ƒ˜ํ”Œ๋ง ์ฃผ๊ธฐ ๋•Œ๋ฌธ์— ์ฃผ์–ด์ง„ ๊ถค์ ์— ๋„๋‹ฌํ•˜๋Š” ์‹œ๊ฐ„์ด ์˜ค๋ž˜ ๊ฑธ๋ฆฐ๋‹ค. ์นด๋ฉ”๋ผ์™€ ์—”์ฝ”๋” ๋ฐ์ดํ„ฐ๋ฅผ ๋™์‹œ์— ์ด์šฉํ•˜๋Š” ๊ฒฝ์šฐ์—๋Š” ์‘๋‹ต์†๋„๊ฐ€ ๋นจ๋ผ์ง€๊ณ  ๋…ธ์ด์ฆˆ๊ฐ€ ์ค„์–ด๋“œ๋Š” ํšจ๊ณผ๊ฐ€ ์žˆ๋Š” ๊ฒƒ์„ ๋ณผ ์ˆ˜ ์žˆ๋‹ค.

๊ทธ๋ฆผ. 8. ๋กœ๋ด‡ ์œ„์น˜์˜ ์ธก์ฒญ ์˜ค์ฐจ ๋ฐ ์ถ”์ • ์˜ค์ฐจ

Fig. 8. Measurement and estimation error

../../Resources/kiee/KIEE.2019.68.5.678/fig8.png

๊ทธ๋ฆผ. 9. ์ด๋™ ๋กœ๋ด‡์˜ ์ถ”์ ์ œ์–ด ์˜ค์ฐจ

Fig. 9. Tracking error of the mobile robot

../../Resources/kiee/KIEE.2019.68.5.678/fig9.png

4.2 ์‹ค์ œ ํ™˜๊ฒฝ ์‹คํ—˜

๋‹ค์ค‘ ์ฃผ๊ธฐ ์ƒ˜ํ”Œ ๋ฐ์ดํ„ฐ ์ƒํƒœ ์ถ”์ •๊ธฐ์™€ ์‹ค์‹œ๊ฐ„ ๋ชจ๋ธ ์˜ˆ์ธก์ œ์–ด๊ธฐ์˜ ์„ฑ๋Šฅ์„ ํ‰๊ฐ€ํ•˜๊ธฐ ์œ„ํ•˜์—ฌ ๋ฌด์ธ ์ง€์ƒ ์ฃผํ–‰ ๋กœ๋ด‡์— ์ ์šฉํ•˜์—ฌ ์‹คํ—˜ํ•˜์˜€๋‹ค. ์—”์ฝ”๋”์˜ ์ƒ˜ํ”Œ๋ง ์ฃผ๊ธฐ๋Š” 0.05์ดˆ ์นด๋ฉ”๋ผ์˜ ์ƒ˜ํ”Œ๋ง ์ฃผ๊ธฐ๋Š” 0.5์ดˆ๋กœ ์ฃผ์–ด์ง„๋‹ค. ๊ทธ๋ฆผ. 10์€ ๊ฐ€๋กœ, ์„ธ๋กœ 1.8m ํฌ๊ธฐ์˜ ์ •์‚ฌ๊ฐํ˜•์˜ ๊ถค์ ์„ ๋”ฐ๋ผ ์ด๋™ํ•˜๋Š” ๊ฒฝ๋กœ ์ถ”์  ์ œ์–ด ์‹คํ—˜ ๊ฒฐ๊ณผ๋ฅผ ๋ณด์—ฌ์ค€๋‹ค.

๊ทธ๋ฆผ. 10. ์ด๋™ ๋กœ๋ด‡์˜ ์ฃผํ–‰ ๊ฒฝ๋กœ

Fig. 10. Trajectory of the mobile robot

../../Resources/kiee/KIEE.2019.68.5.678/fig10.png

ํ‘œ์ง€ํŒ ๊ฒ€์ถœ ์„ฑ๋Šฅ ํ‰๊ฐ€๋ฅผ ์œ„ํ•ด 2๊ฐ€์ง€ ํ‘œ์ง€ํŒ์— ๋Œ€ํ•ด ๊ฐ๊ฐ 20๊ฐœ์˜ ํ…œํ”Œ๋ฆฟ ๊ตฌ์„ฑํ•˜์˜€๊ณ , ์ด 200๊ฐœ์˜ ํ…Œ์ŠคํŠธ ๋ฐ์ดํ„ฐ์— ๋Œ€ํ•ด์„œ ์‹คํ—˜์„ ํ•˜์˜€๋‹ค. ํ‘œ 2์—์„œ ๋‹จ์ผ ํ…œํ”Œ๋ฆฟ๊ณผ ๋‹ค์ค‘ ํ…œํ”Œ๋ฆฟ์€ ๊ธฐ์กด์˜ ํ…œํ”Œ๋ฆฟ ๋งค์นญ ๋ฐฉ๋ฒ•์„ ๋‚˜ํƒ€๋‚ธ๋‹ค. ๋‹จ์ผ ํ…œํ”Œ๋ฆฟ ๋งค์นญ ๋ฐฉ๋ฒ•์—์„œ๋Š” 150$\times$150 ์ด๋ฏธ์ง€ ํ•œ ์žฅ์ด ํ…œํ”Œ๋ฆฟ์œผ๋กœ ์‚ฌ์šฉ๋˜์—ˆ๊ณ  ๋‹ค์ค‘ ํ…œํ”Œ๋ฆฟ ๋งค์นญ ๋ฐฉ๋ฒ•์—์„œ๋Š” 150$\times$150 ์ด๋ฏธ์ง€ 40์žฅ์ด ํ…œํ”Œ๋ฆฟ์œผ๋กœ ์ด์šฉ๋˜์—ˆ๋‹ค. ์ œ์•ˆ๋œ ๋ฐฉ๋ฒ•์—์„œ๋Š” ๋‹ค์ค‘ ํ…œํ”Œ๋ฆฟ ๋งค์นญ์—์„œ ์‚ฌ์šฉ๋˜์–ด์ง„ ํ…œํ”Œ๋ฆฟ์—์„œ ํŠน์ง•์ด ๋˜๋Š” ๋ถ€๋ถ„๋งŒ ์„ ํƒํ•˜์—ฌ ํ…œํ”Œ๋ฆฟ์œผ๋กœ ์ด์šฉํ•˜์˜€๋‹ค(ํฌ๊ธฐ: 50$\times$50). ์ œ์•ˆ๋œ ๋ฐฉ๋ฒ•์€ ๊ธฐ์กด์˜ ๋ฐฉ๋ฒ•๋“ค๋ณด๋‹ค ๋†’์€ ์„ฑ๋Šฅ์„ ๋ณด์˜€๋‹ค.

ํ‘œ 2. ํ‘œ์ง€ํŒ ๊ฒ€์ถœ์„ฑ๋Šฅ ๋น„๊ต

Table 2. Performance comparison of sign detection algorithms

๋‹จ์ผ ํ…œํ”Œ๋ฆฟ

๋‹ค์ค‘ ํ…œํ”Œ๋ฆฟ

์ œ์•ˆ๋œ ๋ฐฉ๋ฒ•

๊ฒ€์ถœ์„ฑ๊ณต

35๊ฐœ

63๊ฐœ

190๊ฐœ

๊ฒ€์ถœ์‹คํŒจ

155๊ฐœ

58๊ฐœ

10๊ฐœ

์˜ค๊ฒ€์ถœ

10๊ฐœ

79๊ฐœ

0๊ฐœ

๊ฒ€์ถœ์„ฑ๋Šฅ

17.5%

31.5%

95%

๊ฒ€์ถœ์‹œ๊ฐ„

0.02s

1.04s

0.31s

๊ทธ๋ฆผ. 11์€ ๋ผ์ด๋”์™€ ์นด๋ฉ”๋ผ ๊ธฐ๋ฐ˜์œผ๋กœ ๊ฒฝ๋กœ๋ฅผ ์ƒ์„ฑํ•˜์—ฌ ์žฅ์• ๋ฌผ์„ ํ”ผํ•ด๊ฐ€๋ฉฐ ์ฃผํ–‰ํ•˜๋Š” ์ž„๋ฌด๋ฅผ ์œ„ํ•œ ์‹คํ—˜ ํ™˜๊ฒฝ์„ ๋ณด์—ฌ์ค€๋‹ค. ์ฃผ์–ด์ง„ ์ด๋™ ๋กœ๋ด‡์˜ ์ž„๋ฌด๋Š” ํ‘œ์ง€ํŒ์ด ๋‚˜์˜ฌ ๋•Œ๊นŒ์ง€ ์žฅ์• ๋ฌผ๊ณผ ์ถฉ๋Œํ•˜์ง€ ์•Š๊ณ  ์ด๋™ํ•ด์•ผํ•˜๋ฉฐ ํ‘œ์ง€ํŒ์ด ๋‚˜์˜ค๋Š” ๊ฒฝ์šฐ ํ‘œ์ง€ํŒ์ด ์•ˆ๋‚ดํ•˜๋Š” ๋ฐฉํ–ฅ์œผ๋กœ ํšŒ์ „์„ ํ•ด์•ผ ํ•œ๋‹ค. ๊ทธ๋ฆผ. 11์—์„œ ๋ณด์—ฌ์ฃผ๋Š” ๋ฐ”์™€ ๊ฐ™์ด ์ œ์•ˆ๋œ ์ž์œจ ์ด๋™๋กœ๋ด‡ ์ถ”์  ์ œ์–ด ์‹œ์Šคํ…œ์„ ํ†ตํ•ด ๋กœ๋ด‡์ด ์„ฑ๊ณต์ ์œผ๋กœ ์ž„๋ฌด๋ฅผ ์ˆ˜ํ–‰ ํ•œ๋‹ค๋Š” ๊ฒƒ์„ ํ™•์ธํ•  ์ˆ˜ ์žˆ๋‹ค.

๊ทธ๋ฆผ. 11. ๋กœ๋ด‡ ์ฃผํ–‰ ์‹คํ—˜ ํ™˜๊ฒฝ ๋ฐ ์ด๋™ ๋กœ๋ด‡์˜ ์ฃผํ–‰ ๊ฒฝ๋กœ

Fig. 11. Experimental setup for driving test and trajectory of the mobile robot

../../Resources/kiee/KIEE.2019.68.5.678/fig11.png

ํ‘œ 3์€ ๋ชจ๋ฐ”์ผ ๋กœ๋ด‡ ์ถ”์ ์ œ์–ด์˜ ํ‰๊ท  ์˜ค์ฐจ๋ฅผ ๋ณด์—ฌ์ค€๋‹ค. ๋กœ๋ด‡์˜ ์œ„์น˜ ์ถ”์  ์˜ค์ฐจ๋Š” ๋‹ค์Œ๊ณผ ๊ฐ™์ด ๊ณ„์‚ฐ๋œ๋‹ค.

ํ‘œ 3. ์ถ”์ ์ œ์–ด์˜ ํ‰๊ท ์˜ค์ฐจ

Table 3. Tracking control mean error

์—”์ฝ”๋”

์นด๋ฉ”๋ผ

์ œ์•ˆ๋œ ๋ฐฉ๋ฒ•

๊ฐ€์ƒํ™˜๊ฒฝ

0.34m

0.22m

0.08m

์‹ค์ œํ™˜๊ฒฝ

0.46m

0.29m

0.16m

$ํ‰๊ท ์˜ค์ฐจ=\dfrac{\sum\sqrt{(ํ˜„์žฌ ๋ชฉํ‘œ์œ„์น˜ - ํ˜„์žฌ๋กœ๋ด‡์œ„์น˜)^{2}}}{์ƒ˜ํ”Œ ์ˆ˜}$

๋ฐ์ดํ„ฐ๋Š” 40์ดˆ ๋™์•ˆ 1์ดˆ๋งˆ๋‹ค ์ธก์ •๋˜์—ˆ์œผ๋ฉฐ (์ƒ˜ํ”Œ ์ˆ˜: 40)ํ˜„์žฌ ๋ชฉํ‘œ ์œ„์น˜์™€ ํ˜„์žฌ ๋กœ๋ด‡ ์œ„์น˜๋Š” ๊ฐ ์ƒ˜ํ”Œ๋งˆ๋‹ค ์ธก์ •๋œ ๋กœ๋ด‡์˜ ์œ„์น˜์™€ ๋ชฉํ‘œ ์œ„์น˜๋ฅผ ๋‚˜ํƒ€๋‚ธ๋‹ค. ์‹คํ—˜์—์„œ ๋กœ๋ด‡์˜ ์ด๋™ ๊ฒฝ๋กœ๋Š” 1.5m*1.5m์˜ ์ •์‚ฌ๊ฐํ˜•์œผ๋กœ ์ฃผ์–ด์ง„๋‹ค. ์—”์ฝ”๋”์˜ ์ƒ˜ํ”Œ๋ง ์ฃผ๊ธฐ๋Š” 0.05์ดˆ ์นด๋ฉ”๋ผ์˜ ์ƒ˜ํ”Œ๋ง ์ฃผ๊ธฐ๋Š” 0.5์ดˆ๋กœ ์ฃผ์–ด์ง„๋‹ค. ๊ฐ€์ƒํ™˜๊ฒฝ ์‹คํ—˜์˜ ๊ฒฝ์šฐ ์ธก์ • ๋…ธ์ด์ฆˆ๋ฅผ ๊ณ ๋ คํ•ด์ฃผ๊ธฐ ์œ„ํ•ด 4.1 ๊ฐ€์ƒํ™˜๊ฒฝ ์‹คํ—˜๊ณผ ๋™์ผํ•œ ์ธก์ • ๋…ธ์ด์ฆˆ ์กฐ๊ฑด์„ ๊ฐ€์ •ํ•˜์˜€๋‹ค.

5. ๊ฒฐ ๋ก 

๋ณธ ๋…ผ๋ฌธ์—์„œ๋Š” ๋‹ค์ค‘ ์„ผ์„œ๋ฅผ ๊ฐ€์ง€๋Š” ์ƒํƒœ์ถ”์ •๊ธฐ ๊ธฐ๋ฐ˜์˜ ์ž์œจ ์ด๋™๋กœ๋ด‡ ์ถ”์ ์ œ์–ด ์‹œ์Šคํ…œ์„ ์ œ์•ˆํ•˜์˜€๋‹ค. ์ œ์•ˆ๋œ ์‹œ์Šคํ…œ์„ ๊ธฐ๋ฐ˜์œผ๋กœ ์‹ค์ œ ๋ฌด์ธ ์ง€์ƒ ๋กœ๋ด‡์— ์ ์šฉํ•˜์˜€์œผ๋ฉฐ ์‹คํ—˜ ๊ฒฐ๊ณผ๋ฅผ ํ†ตํ•˜์—ฌ ๋‹ค์ค‘ ์„ผ์„œ ๊ธฐ๋ฐ˜์˜ ์ž์œจ ์ถ”์  ์ œ์–ด ์‹œ์Šคํ…œ์˜ ์žฅ์ ์„ ๊ฒ€์ฆํ•˜์˜€๋‹ค.

์ถ”ํ›„ ์—ฐ๊ตฌ ๊ณ„ํš์œผ๋กœ ์นด๋ฉ”๋ผ, ์—”์ฝ”๋” ๋ฟ๋งŒ ์•„๋‹ˆ๋ผ ๋ผ์ด๋‹ค, ๊ด€์„ฑ ์ธก์ • ์žฅ์น˜ (IMU) ๋“ฑ๊ณผ ๊ฐ™์€ ๋‹ค์–‘ํ•œ ์„ผ์„œ๋“ค์„ ์ด์šฉํ•˜์—ฌ ์—ฐ๊ตฌ ๊ฒฐ๊ณผ๋ฅผ ํ™•์žฅํ•  ๊ฒƒ์ด๋‹ค. ๋˜ํ•œ ๋” ํ˜„์‹ค์ ์ธ ์ง€๋Šฅ ์‹œ์Šคํ…œ ๊ตฌํ˜„์„ ์œ„ํ•˜์—ฌ ๋‹ค์ค‘ ์„ผ์„œ ๊ธฐ๋ฐ˜์˜ ๊ฒฝ๋กœ ์ƒ์„ฑ ์•Œ๊ณ ๋ฆฌ์ฆ˜์ด ์—ฐ๊ตฌ ๋  ๊ฒƒ์ด๋‹ค.

Acknowledgements

This research was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and Technology (2016R1D1A1B03930623).

References

1 
Moshe Kam, Xiaoxun Zhu, Paul Kalata, 1997, Sensor fusion for mobile robot navigation, Proceedings of the IEEE, Vol. 85, No. 1, pp. 108-119Google Search
2 
Derradji Nada, Mounir Bousbia-Salah, Maamar Bettayeb, 2018, Multi-sensor data fusion for wheelchair position estimation with unscented Kalman Filter, International Journal of Automation and Computing, Vol. 15, No. 2, pp. 207-217DOI
3 
Yashar Balazadegan Sarvrood, Siavash Hosseinyalamdary, Yang Gao, 2016, Visual-LiDAR odometry aided by reduced IMU, ISPRS International Journal of Geo-Information, Vol. 5, No. , pp. 1-24DOI
4 
Rong-Jong Wai, You-Wei Lin, 2013, Adaptive Moving-Target Tracking Control of a Vision-Based Mobile Robot via a Dynamic Petri Recurrent Fuzzy Neural Network, IEEE Trans. Fuzzy Systems, Vol. 21, No. 4, pp. 688-701DOI
5 
David Tick, et al, 2013, Tracking control of mobile robots localized via chained fusion of discrete and continuous epipolar geometry, IMU and odometry, IEEE transactions on cybernetics, Vol. 43, No. 4, pp. 1237-1250DOI
6 
Marcus Hebel, Uwe Stilla, 2017, LiDAR-supported navigation of UAVs over urban areas, Surveying and Land Information Science, Vol. 70, No. 3, pp. 139-149Google Search
7 
Shu-Li Sun, Zi-Li Deng, 2004, Multi-sensor optimal information fusion Kalman filter, Automatica, Vol. 40, No. 6, pp. 1017-1023DOI
8 
Bรกlint Takรกcs, et al, 2016, Python code generation for explicit MPC in MPT, 2016 Eurooean Control conference (ECC), pp. 1328-1333DOI
9 
Niko Sรผnderhauf, Peter Protzel, 2007, Stereo odometry โ€” a review of approaches, Chemnitz University of Technology Technical ReportGoogle Search
10 
Seungyong Han, Sangmoon Lee, 2018, Sampled-Data MPC for Leader-Following of Multi-Mobile Robot System, European Journal of Control, Vol. 67, No. 2, pp. 308-313DOI
11 
Miad Moarref, Luis Rodrigues, 2014, Observer design for linear multi-rate sampled-data systems, 2014 American Control Conference (ACC), pp. 5319-5324DOI

์ €์ž์†Œ๊ฐœ

์ง„ ์šฉ ์‹ (Yongsik Jin)
../../Resources/kiee/KIEE.2019.68.5.678/au1.png

He is currently toward the Ph.D. degree with the School of Electronics Engineering, Kyungpook National University, Republic of Korea. He is with the Cyber Physical Systems & Control Lab, Kyungpook National University. His current research interests include deep neural networks, autonomous learning systems, cyber physical system, and networked control systems.

E-mail : yongsik@knu.ac.kr

ํ•œ ์Šน ์šฉ (Seungyong Han)
../../Resources/kiee/KIEE.2019.68.5.678/au2.png

He is currently toward the Ph.D. degree with the School of Electronics Engineering, Kyungpook National University, Republic of Korea. He is with the Cyber Physical Systems & Control Lab, Kyungpook National University. His current research interests include Networked control systems, model predictive control, and autonomous vehicle control systems.

E-mail : seungyong@knu.ac.kr

์ด ์ƒ ๋ฌธ (Sangmoon Lee)
../../Resources/kiee/KIEE.2019.68.5.678/au3.png

Sangmoon Lee received the B.S. degree in electronics engineering from Kyungpook National University, Daegu, South Korea, in 1999, the M.S and Ph.D. degrees in electronics engineering from POSTECH, Pohang, South Korea, in 2001 and 2006, respectively., He is currently an Associate Professor with the School of Electronics Engineering, Kyungpook National University. His main research interests include cyber physical systems control, networked control systems, nonlinear systems, fuzzy systems, robust control, model predictive control, and its industrial applications.

E-mail : moony@knu.ac.kr