MAapping Robotics System (MARS)

Our group works on 3D urban mapping using cameras, LiDARs, and other inertial sensors. We use two different sensor systems for mapping, depending on the LiDAR configuration, 1) push broom style and 2) 360 tilted scanning style.


[Fig] Sensor specification

System Sensors

For more detailed explanation for each sensor system and comparison, please refer to our URAI 2017 paper.

Joowan Kim, Jinyong Jeong, Young-Sik Shin, Younggun Cho, Hyunchul Roh and Ayoung Kim, "LiDAR Configuration Comparison for Urban Mapping System", URAI 2017.

Some research achievement we obtained so far is at 3D urban mapping and lane map generation.

1) 3D Map of Urban Canyon Urban 3D map accuracy and consistency heavily rely on the localization accuracy of the mapping vehicle. The vehicle localization is challenging in an urban area where high rise building causes sporadic GPS reception. Even if the GPS is received, the consumer level GPS can merely guarantee meter-level accuracy. An initial and fast solution we suggest is to leverage the existing 3D map and turn them into a dense 3D map. For this research, our focus is at automating the mapping procedure based upon Simultaneous Localization and Mapping (SLAM) technique.

Hyunchul Roh and Jinyong Jeong and Younggun Cho and Ayoung Kim, Accurate Mobile Urban Mapping via Digital Map-Based SLAM (Sensors 2016)

Another research focus we aim is to correct vehicle's navigational error effectively within the urban environment. We noted the two things. First, the GPS signal and structural information from buildings are complementary. When buildings are around GPS becomes unreliable; GPS reception is clear in an open space with no buildings around. Secondly, the most critical error is the heading error rather than positional error when targeting a large scale mapping. Based on these two ideas, we introduce to correct the directional error from building information and aerial images.

Click to see video

Hyunchul Roh and Jinyong Jeong and Ayoung Kim, Aerial Image based Heading Correction for Large Scale SLAM in an Urban Canyon (IROS 2017 with RA-L)

2) Lane map for autonomous vehicles Generation of accurate lane map is a challenge in terms of the vehicle localization. Using a monocular camera for lane map generation is achieved. Furthermore, we present using the generated lane maps for vehicle localization.

Click to see video

Jinyong Jeong and Younggun Cho and Ayoung Kim, Road-SLAM: Road Marking based SLAM with Lane-level Accuracy (IV 2017)

Related Projects

  • [KEIT 2015 - 2020] Collaboration with prof. Hyunchul Shim (PI)
    Development of real-time localization and 3-D mapping system with cm-level accuracy based on digital maps and vision data for autonomous driving
    (자율 주행을 위한 영상 및 전자지도 기반 실시간 정밀 측위 (오차 10cm미만) 및 3차원 지도 생성 기술 개발)
  • [NAVER, NAVERLABS 2016-2017]
    도심 환경 지도 제작 산학 협력
Copyright © 2014-2019 IRAP
Page last modified on July 30, 2017, at 10:30 PM