Simultaneous localization and mapping (SLAM) for automotive using forward looking radar

Localization is the process of finding ones relation to the surrounding stationary objects, and mapping is the process of determining the relation between the stationary objects. Mapping requires a sensing technique, and in addition to that a known (or estimated) location. If we neglect the possibility of external support such as GPS or street maps and instead consider localization using ranging sensors, we are facing two strongly interconnected problems that needs to be solved simultaneously. We have investigated the possibility to perform SLAM (simultaneous localization and mapping) in automotive using a forward-looking radar, primarily designed for ACC (adaptive cruise control) and PCS (pre-crash safety). Our positioning is not only relying on the inertial measurements speed and yaw-rate from the vehicle, but also incorporates radar measurements and performs extended Kalman filter SLAM (EKF-SLAM). Using collected data from a single forward-looking radar, we have shown that it is possible to enhance the positioning performance without support from GPS. The heading of the vehicle was drifting as yaw-rate error accumulated, but when adding EKF-SLAM we mitigated this problem. However, this system was very sensitive to parameter settings, as well as radar misalignment, and needs a thorough on-line calibration. Robustness can be achieved by increasing the field of view or having side-looking radars.

Language

  • English

Media Info

  • Pagination: pp 258-265
  • Monograph Title: FAST-zero'15: 3rd international symposium on future active safety technology toward zero traffic accidents: September 9-11, 2015 Gothenburg, Sweden: proceedings

Subject/Index Terms

Filing Info

  • Accession Number: 01602293
  • Record Type: Publication
  • Source Agency: Swedish National Road and Transport Research Institute (VTI)
  • Files: ITRD, VTI
  • Created Date: Jun 20 2016 1:28PM