Tsukada Lab. Computer Network Group

Localizability Estimation for Autonomous Driving: A Deep Learning-Based Feature Extraction Approach

Kazuto Matsumoto

Master's thesis, Graduate School of Information Science and Technology, The University of Tokyo, Mar 30, 2024

Summary in three lines

  • LiDAR map matching can lose accuracy in places with few distinctive features.
  • The thesis uses deep learning to extract features from vehicle LiDAR and point-cloud maps and predict localization accuracy.
  • It compares point-cloud representations using MinkLoc3D and MinkLoc3D-SI.

Figures from the paper

Figure 3.3: Localization accuracy prediction using map and LiDAR features
Figure 3.3. Localization accuracy prediction using map and LiDAR featuresSource: Figure 3.3 from Kazuto Matsumoto, Master's thesis, Graduate School of Information Science and Technology, The University of Tokyo, 2024 (original artwork from the authors’ manuscript)Map and vehicle point clouds contribute features to the model that estimates localization errors.

How to cite

K. Matsumoto, “Localizability Estimation for Autonomous Driving: A Deep Learning-Based Feature Extraction Approach,” in Master's thesis, Graduate School of Information Science and Technology, The University of Tokyo, 2024.

BibTeX

@mastersthesis{Matsumoto2024Thesis,
  title     = {Localizability Estimation for Autonomous Driving: A Deep Learning-Based Feature Extraction Approach},
  author    = {Kazuto Matsumoto},
  year      = {2024},
  date      = {2024-03-30},
  urldate   = {2024-03-31},
  school    = {Master's thesis, Graduate School of Information Science and Technology, The University of Tokyo},
  keywords  = {ITS, Thesis},
  pubstate  = {published},
  tppubtype = {mastersthesis}
}