Trend AnalysisOther Engineering
Autonomous Vehicles Underground: Navigating Mines Without GPS
Underground mines are among the most challenging environments for autonomous vehicles: no GPS, limited visibility, narrow tunnels, and hazardous conditions. Recent advances in LiDAR-based localization, behavior-based control, and post-blast UAV inspection are making autonomous mining operations increasingly practical.
By Sean K.S. Shin
This blog summarizes research trends based on published paper abstracts. Specific numbers or findings may contain inaccuracies. For scholarly rigor, always consult the original papers cited in each post.
Underground mining is one of the most dangerous occupations globallyโhazards include rock falls, toxic gases, flooding, equipment collisions, and blasting accidents. Autonomous vehicles could significantly reduce human exposure to these risks by performing haulage, inspection, and rescue operations without human operators in the most dangerous areas. But underground environments present technical challenges that surface autonomous vehicles do not face: no GPS signals, limited or absent lighting, dust and particulates that degrade sensors, narrow and irregular tunnel geometries, and the need to operate safely alongside human workers.
The Research Landscape
Post-Blast UAV Inspection
Nordstrรถm and Dahlquist (2025), with 5 citations, report what they describe as the first fully autonomous UAV mission to perform gas measurements after a real blast in an underground mine. The mission was deployed approximately 40 minutes after blastingโa period when the area is hazardous due to residual gases (CO, NOโ, SOโ) and unstable rock.
The UAV navigated autonomously through the mine using LiDAR-based SLAM (Simultaneous Localization and Mapping), measured gas concentrations at multiple points, and returned safely. The practical value is clear: currently, human workers must wait hours before entering post-blast areas for gas clearance. Autonomous UAV inspection can perform this assessment much soonerโpotentially allowing production to resume faster while keeping workers safe.
Behavior-Based Navigation for Narrow Tunnels
Badr and Almaghout (2024), with 1 citation, address the specific challenge of autonomous navigation in narrow mine tunnels where conventional path planning algorithms struggle. Their behavior-based control approach draws on robotics techniques that decompose complex navigation into simple reactive behaviors (wall-following, obstacle avoidance, goal-seeking) that combine to produce robust tunnel navigation.
The approach is notable for its robustness: it handles irregular tunnel geometries, dynamic obstacles (fallen rock, other vehicles), and sensor degradation (dust, water on LiDAR lenses) better than optimization-based planners that assume accurate environmental models.
Articulated LHD Automation
Wu and Lu (2025) present an autonomous driving system for articulated Load-Haul-Dump (LHD) machinesโlarge vehicles that scoop ore and transport it through tunnels. LHDs present additional challenges: they are articulated (bending in the middle), non-holonomic (cannot move sideways), and operate in feature-poor environments where LiDAR scans show little variation between one section of tunnel and the next.
Their system was validated in a real underground mine, demonstrating autonomous loading, hauling, and dumping cyclesโthe core operational workflow that LHDs perform continuously during production.
Cyber-Physical Safety System
Behera, Agarwal, and Badr & Almaghout (2024), with 2 citations, take a systems approach: building a cyber-physical system (CPS) that integrates an unmanned ground vehicle with environmental mapping, gas sensing, and computational intelligence for mine safety and rescue support. The UGV maps unknown mine sections, identifies hazards, and communicates findings to surface operators through a real-time data link.
Critical Analysis: Claims and Evidence
<
| Claim | Evidence | Verdict |
|---|
| Autonomous UAVs can safely inspect post-blast mine areas | Nordstrรถm et al.'s real-mine demonstration | โ
Supported โ first real-world deployment |
| Behavior-based control handles narrow, irregular mine tunnels | Badr & Almaghout's simulation and limited field tests | โ
Supported โ robustness demonstrated |
| Autonomous LHD haulage is feasible in real underground mines | Wu & Lu's mine validation | โ
Supported โ complete load-haul-dump cycles |
| CPS-based UGVs can provide mine safety and rescue support | Behera et al.'s prototype system | โ ๏ธ Uncertain โ prototype demonstrated; operational deployment not yet |
What This Means for Your Research
For mining engineers, autonomous vehicles are moving from research prototypes to operational systems. The post-blast UAV inspection use case (Nordstrรถm et al.) is likely the nearest to commercial deployment. For robotics researchers, underground mines provide a uniquely challenging testbed for autonomous navigation.
Explore related work through ORAA ResearchBrain.
๋ฉด์ฑ
์กฐํญ: ์ด ๊ฒ์๋ฌผ์ ์ ๋ณด ์ ๊ณต์ ์ํ ์ฐ๊ตฌ ๋ํฅ ๊ฐ์์ด๋ค. ํ์ ์ฐ๊ตฌ์์ ์ธ์ฉํ๊ธฐ ์ ์ ํน์ ์ฐ๊ตฌ ๊ฒฐ๊ณผ, ํต๊ณ ๋ฐ ์ฃผ์ฅ์ ์๋ณธ ๋
ผ๋ฌธ์ ํตํด ๋ฐ๋์ ๊ฒ์ฆํด์ผ ํ๋ค.
์งํ์ ์์จ์ฃผํ ์ฐจ๋: GPS ์์ด ๊ด์ฐ์ ํญ๋ฒํ๋ค
์งํ ๊ด์ฐ ์ฑ๊ตด์ ์ ์ธ๊ณ์ ์ผ๋ก ๊ฐ์ฅ ์ํํ ์ง์
์ค ํ๋์ด๋คโ๋์, ์ ๋
๊ฐ์ค, ์นจ์, ์ฅ๋น ์ถฉ๋, ๋ฐํ ์ฌ๊ณ ๋ฑ์ ์ํ์ด ์กด์ฌํ๋ค. ์์จ์ฃผํ ์ฐจ๋์ ๊ฐ์ฅ ์ํํ ๊ตฌ์ญ์์ ์ธ๊ฐ ์ด์ ์ ์์ด ์ด๋ฐ, ์ ๊ฒ, ๊ตฌ์กฐ ์์
์ ์ํํจ์ผ๋ก์จ ์ธ๊ฐ์ด ์ด๋ฌํ ์ํ์ ๋
ธ์ถ๋๋ ๊ฒ์ ํฌ๊ฒ ์ค์ผ ์ ์๋ค. ๊ทธ๋ฌ๋ ์งํ ํ๊ฒฝ์ ์ง์ ์์จ์ฃผํ ์ฐจ๋์ด ์ง๋ฉดํ์ง ์๋ ๊ธฐ์ ์ ๊ณผ์ ๋ฅผ ์๊ณ ์๋ค: GPS ์ ํธ ๋ถ์ฌ, ์ ํ์ ์ด๊ฑฐ๋ ์์ ํ ์๋ ์กฐ๋ช
, ์ผ์๋ฅผ ์ ํ์ํค๋ ๋จผ์ง์ ๋ฏธ๋ฆฝ์, ์ข๊ณ ๋ถ๊ท์นํ ํฐ๋ ํ์, ๊ทธ๋ฆฌ๊ณ ์ธ๊ฐ ์์
์์ ํจ๊ป ์์ ํ๊ฒ ์ด์ฉ๋์ด์ผ ํ ํ์์ฑ์ด ๊ทธ๊ฒ์ด๋ค.
์ฐ๊ตฌ ํํฉ
๋ฐํ ํ UAV ์ ๊ฒ
Nordstrรถm๊ณผ Dahlquist(2025)๋ 5ํ ์ธ์ฉ์ ๊ธฐ๋กํ๋ฉฐ, ์ค์ ์งํ ๊ด์ฐ์์ ๋ฐํ ํ ๊ฐ์ค ์ธก์ ์ ์ํํ ์ต์ด์ ์์ ์์จ UAV ์๋ฌด๋ผ๊ณ ์ค๋ช
ํ๋ ๋ด์ฉ์ ๋ณด๊ณ ํ๋ค. ํด๋น ์๋ฌด๋ ๋ฐํ ์ฝ 40๋ถ ํ์ ํฌ์
๋์๋๋ฐ, ์ด ์๊ธฐ๋ ์๋ฅ ๊ฐ์ค(CO, NOโ, SOโ)์ ๋ถ์์ ํ ์์์ผ๋ก ์ธํด ํด๋น ๊ตฌ์ญ์ด ์ํํ ์ํ์ ์๋ ๋์ด๋ค.
UAV๋ LiDAR ๊ธฐ๋ฐ SLAM(๋์ ์์น ์ถ์ ๋ฐ ์ง๋ ์์ฑ)์ ์ฌ์ฉํ์ฌ ๊ด์ฐ ๋ด๋ฅผ ์์จ์ ์ผ๋ก ํญ๋ฒํ๊ณ , ์ฌ๋ฌ ์ง์ ์์ ๊ฐ์ค ๋๋๋ฅผ ์ธก์ ํ ํ ์์ ํ๊ฒ ๊ทํํ์๋ค. ์ค์ฉ์ ๊ฐ์น๋ ๋ช
ํํ๋ค: ํ์ฌ ์ธ๊ฐ ์์
์๋ ๊ฐ์ค ์ ๊ฑฐ๋ฅผ ์ํด ๋ฐํ ํ ๋ช ์๊ฐ์ด ์ง๋์ผ ํด๋น ๊ตฌ์ญ์ ์ง์
ํ ์ ์๋ค. ์์จ UAV ์ ๊ฒ์ ์ด ํ๊ฐ๋ฅผ ํจ์ฌ ๋ ๋นจ๋ฆฌ ์ํํ ์ ์์ดโ์ ์ฌ์ ์ผ๋ก ์์
์๋ฅผ ์์ ํ๊ฒ ๋ณดํธํ๋ฉด์ ์์ฐ์ ๋ ๋น ๋ฅด๊ฒ ์ฌ๊ฐํ ์ ์๋ค.
์ข์ ํฐ๋์ ์ํ ํ๋ ๊ธฐ๋ฐ ํญ๋ฒ
Badr์ Almaghout(2024)์ 1ํ ์ธ์ฉ์ ๊ธฐ๋กํ๋ฉฐ, ๊ธฐ์กด์ ๊ฒฝ๋ก ๊ณํ ์๊ณ ๋ฆฌ์ฆ์ด ์ด๋ ค์์ ๊ฒช๋ ์ข์ ๊ด์ฐ ํฐ๋์์์ ์์จ ํญ๋ฒ์ด๋ผ๋ ๊ตฌ์ฒด์ ์ธ ๊ณผ์ ๋ฅผ ๋ค๋ฃฌ๋ค. ์ด๋ค์ ํ๋ ๊ธฐ๋ฐ ์ ์ด ์ ๊ทผ๋ฒ์ ๋ณต์กํ ํญ๋ฒ์ ๋จ์ํ ๋ฐ์์ ํ๋(๋ฒฝ ์ถ์ข
, ์ฅ์ ๋ฌผ ํํผ, ๋ชฉํ ํ์)์ผ๋ก ๋ถํดํ์ฌ ์กฐํฉํจ์ผ๋ก์จ ๊ฐ๊ฑดํ ํฐ๋ ํญ๋ฒ์ ๊ตฌํํ๋ ๋ก๋ด๊ณตํ ๊ธฐ๋ฒ์ ํ์ฉํ๋ค.
์ด ์ ๊ทผ๋ฒ์ ๊ฐ๊ฑด์ฑ ๋ฉด์์ ์ฃผ๋ชฉํ ๋งํ๋ค: ์ ํํ ํ๊ฒฝ ๋ชจ๋ธ์ ๊ฐ์ ํ๋ ์ต์ ํ ๊ธฐ๋ฐ ๊ณํ๊ธฐ๋ณด๋ค ๋ถ๊ท์นํ ํฐ๋ ํ์, ๋์ ์ฅ์ ๋ฌผ(๋์, ๋ค๋ฅธ ์ฐจ๋), ๊ทธ๋ฆฌ๊ณ ์ผ์ ์ ํ(๋จผ์ง, LiDAR ๋ ์ฆ์ ์๋ถ)์ ๋ ์ ๋์ฒํ๋ค.
๊ตด์ ์ LHD ์๋ํ
Wu์ Lu(2025)๋ ๊ตด์ ์ LHD(Load-Haul-Dump) ๊ธฐ๊ณโ๊ด์์ ํผ์ฌ๋ ค ํฐ๋์ ํตํด ์ด๋ฐํ๋ ๋ํ ์ฐจ๋โ๋ฅผ ์ํ ์์จ์ฃผํ ์์คํ
์ ์ ์ํ๋ค. LHD๋ ์ถ๊ฐ์ ์ธ ๊ณผ์ ๋ฅผ ์๊ณ ์๋ค: ๊ตด์ ์(์ค๊ฐ์์ ๊บพ์ด๋ ๊ตฌ์กฐ)์ด๊ณ , ๋นํ๋ก๋
ธ๋ฏน(์ธก๋ฉด ์ด๋ ๋ถ๊ฐ)์ด๋ฉฐ, ํฐ๋์ ํ ๊ตฌ๊ฐ๊ณผ ๋ค์ ๊ตฌ๊ฐ ์ฌ์ด์ LiDAR ์ค์บ์ด ๊ฑฐ์ ๋ณํ๋ฅผ ๋ณด์ด์ง ์๋ ํน์ง์ด ๋น์ฝํ ํ๊ฒฝ์์ ์ด์ฉ๋๋ค.
์ด๋ค์ ์์คํ
์ ์ค์ ์งํ ๊ด์ฐ์์ ๊ฒ์ฆ๋์์ผ๋ฉฐ, ์์จ ์ ์ฌ, ์ด๋ฐ, ํ์ญ ์ฃผ๊ธฐโ์์ฐ ์ค LHD๊ฐ ์ง์์ ์ผ๋ก ์ํํ๋ ํต์ฌ ์ด์ฉ ์์
ํ๋ฆโ๋ฅผ ์์ฐํ์๋ค.
์ฌ์ด๋ฒ-๋ฌผ๋ฆฌ ์์ ์์คํ
Behera, Agarwal, ๊ทธ๋ฆฌ๊ณ Badr์ Almaghout(2024)์ 2ํ ์ธ์ฉ์ ๊ธฐ๋กํ๋ฉฐ, ์์คํ
์ ์ ๊ทผ๋ฒ์ ์ทจํ๋ค: ๋ฌด์ธ ์ง์ ์ฐจ๋(UGV)์ ํ๊ฒฝ ์ง๋ ์์ฑ, ๊ฐ์ค ๊ฐ์ง, ๊ทธ๋ฆฌ๊ณ ๊ด์ฐ ์์ ๋ฐ ๊ตฌ์กฐ ์ง์์ ์ํ ๊ณ์ฐ ์ง๋ฅ๊ณผ ํตํฉํ๋ ์ฌ์ด๋ฒ-๋ฌผ๋ฆฌ ์์คํ
(CPS)์ ๊ตฌ์ถํ๋ค. UGV๋ ๋ฏธ์ง์ ๊ด์ฐ ๊ตฌ์ญ์ ์ง๋ํํ๊ณ , ์ํ ์์๋ฅผ ์๋ณํ๋ฉฐ, ์ค์๊ฐ ๋ฐ์ดํฐ ๋งํฌ๋ฅผ ํตํด ์ง์ ์ด์ฉ์์๊ฒ ๊ฒฐ๊ณผ๋ฅผ ์ ๋ฌํ๋ค.
๋นํ์ ๋ถ์: ์ฃผ์ฅ๊ณผ ๊ทผ๊ฑฐ
<
| ์ฃผ์ฅ | ๊ทผ๊ฑฐ | ํ์ |
|---|
| ์์จ UAV๊ฐ ๋ฐํ ํ ๊ด์ฐ ๊ตฌ์ญ์ ์์ ํ๊ฒ ์ ๊ฒํ ์ ์๋ค | Nordstrรถm ๋ฑ์ ์ค์ ๊ด์ฐ ์์ฐ | โ
์ง์ง๋จ โ ์ต์ด์ ์ค์ธ๊ณ ๋ฐฐ์น |
| Behavior-based control handles narrow, irregular mine tunnels | Badr & Almaghout's simulation and limited field tests | โ
Supported โ robustness demonstrated |
| Autonomous LHD haulage is feasible in real underground mines | Wu & Lu's mine validation | โ
Supported โ complete load-haul-dump cycles |
| CPS-based UGVs can provide mine safety and rescue support | Behera et al.'s prototype system | โ ๏ธ Uncertain โ prototype demonstrated; operational deployment not yet |
What This Means for Your Research
For mining engineers, autonomous vehicles are moving from research prototypes to operational systems. The post-blast UAV inspection use case (Nordstrรถm et al.) is likely the nearest to commercial deployment. For robotics researchers, underground mines provide a uniquely challenging testbed for autonomous navigation.
Explore related work through ORAA ResearchBrain.
References (5)
[1] Nordstrรถm, S., Stathoulopoulos, N., & Dahlquist, N. (2025). Safety Inspections and Gas Monitoring in Hazardous Mining Areas Shortly After Blasting Using Autonomous UAVs. Journal of Field Robotics.
[2] Badr, A.A. & Almaghout, K. (2024). Navigating Narrow Margins: Behavior-Based Control for Autonomous Mining Vehicles. Regular and Chaotic Dynamics.
[3] Wu, K. & Lu, Z. (2025). An Autonomous Driving System for Articulated Underground LHD Machines. Proc. RICAI 2025, IEEE.
[4] Behera, L., Agarwal, S., & Sandhan, T. (2025). A cyber-physical system based UGV for safety inspection and rescue support in underground mine. International Journal of Intelligent Unmanned Systems.
Behera, L., Agarwal, S., Sandhan, T., Sharma, P., Kumar, A., Ranjan, A., et al. (2025). A cyber-physical system based unmanned ground vehicles for safety inspection and rescue support inย anย underground mine. International Journal of Intelligent Unmanned Systems, 13(1), 92-128.