Fusion SLAM
Home
  • OVERVIEW AND INTRODUCTION
  • INSTALLING
  • DATASETS
  • TUTORIALS FOR ROVIOLI (ONLINE FRONTEND)
  • TUTORIALS FOR USING MAPLAB (OFFLINE TOOLS) - BASICS
  • TUTORIALS FOR USING MAPLAB (OFFLINE TOOLS) - USE-CASES
  • TUTORIALS FOR USING MAPLAB SERVER (ONLINE)
  • HARDWARE INTEGRATION AND SENSOR CALIBRATION
  • TUTORIALS FOR EXTENDING MAPLAB
  • DEVELOPEMENT GUIDELINES
  • ADDITIONAL FORMATS
Project
  • 简体中文
  • English
Home
  • OVERVIEW AND INTRODUCTION
  • INSTALLING
  • DATASETS
  • TUTORIALS FOR ROVIOLI (ONLINE FRONTEND)
  • TUTORIALS FOR USING MAPLAB (OFFLINE TOOLS) - BASICS
  • TUTORIALS FOR USING MAPLAB (OFFLINE TOOLS) - USE-CASES
  • TUTORIALS FOR USING MAPLAB SERVER (ONLINE)
  • HARDWARE INTEGRATION AND SENSOR CALIBRATION
  • TUTORIALS FOR EXTENDING MAPLAB
  • DEVELOPEMENT GUIDELINES
  • ADDITIONAL FORMATS
Project
  • 简体中文
  • English
  • OVERVIEW AND INTRODUCTION

    • Introduction to the Maplab Framework
    • Main Papers
    • Additional Citations
    • Related Research
    • FAQ
    • Known Issues
  • INSTALLING

    • Installing on Ubuntu
    • Compilation and Debugging
  • DATASETS
  • TUTORIALS FOR ROVIOLI (ONLINE FRONTEND)

    • ROVIOLI Introduction
    • Running ROVIOLI in VIO mode: calibartion files, rostopics, bag/topic mode, visualization
    • Running ROVIOLI in Localization mode
    • Multi-session mapping with ROVIOLI
  • TUTORIALS FOR USING MAPLAB (OFFLINE TOOLS) - BASICS

    • Basic Console Usage
    • Parameters (Gflags)
    • Console map management: load, save, basic visualization
    • Inspecting and visualizing a map
    • Map visualization: see your map in RViz!
    • Preparing a single session map: optimization, loop-closure
    • Understanding loop-closure
    • Optimizing VI-Maps
    • Preparing a multi-session map: map anchoring, loop-closure, pose-graph relaxation
    • Dense Reconstruction: attaching resources to map, available reconstruction tools
    • Resource Importer
  • TUTORIALS FOR USING MAPLAB (OFFLINE TOOLS) - USE-CASES

    • Multi-session use case: CLA, multi-floor use-case
    • Map sparsification: make your mapping more efficient
    • Stereo Dense Reconstruction: EuRoC, multi-session reconstruction use-case
    • External Features
  • TUTORIALS FOR USING MAPLAB SERVER (ONLINE)
  • HARDWARE INTEGRATION AND SENSOR CALIBRATION

    • Sensor Calibration Format: ncamera, imu-sigmas
    • Initial sensor calibration with Kalibr
    • Sensor calibration refinement
    • Intel RealSense ZR300
    • VersaVIS
  • TUTORIALS FOR EXTENDING MAPLAB

    • Using the MapManager
    • Using Timing and Statistics
    • /maplab/docs/pages/tutorials-extending-maplab/C_Coding-Examples:-Creating-a-custom-console-plugin.html
    • /maplab/docs/pages/tutorials-extending-maplab/D_Coding-Examples:-Working-with-the-VI-Map.html
    • Console Plugin System
  • DEVELOPEMENT GUIDELINES

    • Importing maplab to Eclipse
    • Contributing to maplab
    • Header Include Guide
    • Debugging applications
    • Expressing frame transformations in code
    • Verbosity Policy
  • ADDITIONAL FORMATS

Optimizing VI-Maps

Commands:

optv     # Optimization based on the visual data only (landmark observations)
optvi    # Optimization based on the visual and inertial terms (landmarks and IMU)

Hints:

  • The optimization can be aborted at any time using Ctrl-C, it will then abort after at the end of the current iteration.
  • Important flags:
    --ba_num_iterations                # Set the maximum number of iterations of the optimization.
    --ba_visualize_every_n_iterations  # Visualize the result of the optimization at every Nth step.
    
  • Only landmarks flagged as 'good' will be part of the optimization. The following flags can be used to set the parameters of the quality metrics:
    --elq_min_observation_angle_deg
         # Minimum angle disparity of observers for a landmark to be well constrained.
    --elq_min_observers
         # Minimum number of observers for a landmark to be well constrained.
    --elq_min_distance_from_closest_observer
         # A landmark needs to be at least as far away from the observer to be well constrained [m].
    --elq_max_distance_from_closest_observer
         # A landmark cannot be further away from the observer than this to be well constrained [m].
    
    The landmark quality can be (re-)evaluated using the rtl or 'evaluate_landmark_quality command (before the optimization):
    evaluate_landmark_quality  --elq_min_observers=2 ...
    
Prev
Understanding loop-closure
Next
Preparing a multi-session map: map anchoring, loop-closure, pose-graph relaxation