Skip to content

Latest commit

 

History

History
27 lines (25 loc) · 1.51 KB

gvins.md

File metadata and controls

27 lines (25 loc) · 1.51 KB

GVINS: Tightly Coupled GNSS–Visual–Inertial Fusion for Smooth and Consistent State Estimation

Author : Cao Year : 2022

Notes:

  • Fusion of VIO with GNSS pseudo-range and doppler shift measurements in a factor graph
  • a quantity tought to be estimated is the 4 Dof transform between the local VIN frame and the global GNSS frame
  • tight fusion
  • GNSS - VI fusion with EKF : UAV position and attitude estimation using IMU, GNSS and camera + A robust and modular multi-sensor fusion approach applied to MAV navigation
  • ECEF = earth centred, earth fixed $xy$ coincide with earth equatorial plane with $z$ pointing to the north and $x$ to the prime meridian
  • ENU = semi global frame that is East, North and Up (gravity aligned)
  • ECI = earth centred inertial frame: inertial coordinate system with center of mass of earth as origin and fixed wrt the stars (ie does not rotate with earth)
  • Estimates a sliding window of poses and inverse depth, the yaw offset between local world frame and ENU frame, receiver clock bias and receiver drifting rate
  • "pseudo" range cause contains various errors
  • Single Point Positionning (SPP) algo to perform trilateration in the ECI frame => four satellites needed
  • INIT :
    • classic Sfm for VI init
    • coarse anchor pose with SPP
    • yaw alignment with doppler meas
    • anchor refinement that imposes a clock constraint
  • Uses Doppler to find yaw: have a velocity estimate to align with VIO one
  • Yaw offset cannot be estimated with low velocity
  • loop closure disabled for experiments