Autonomous cars use GPS to perform coarse location estimation using a method called trilateration. However, the precision is 1 to 10 m, which is dangerous. Hence, localization needs to be done

Localization methods

  • Odometry - Use of data from motion sensors (starting position and wheel displacement) to estimate a position. Generally inaccurate
  • Kalman Filters - State estimation of vehicle. Combines noisy and inaccurate data together to produce accurate data
  • Particle Filters - Variant of Bayesian Filters. Creates particles around areas where observations are similar to maps
  • SLAM - Simultaneous Localization And Mapping. Estimation of position and landmarks