The Future of SLAM

Posted on Edited on In ,

Talks about the future development directions of SLAM.

Directions

  • Hardware acceleration
    • Use GPU and framework to accelerate the calculation, e.g. feature extraction and matching, BA solving, to improve the efficiency of the algorithm
  • New sensor & Sensor fusion
    • Fisheye camera
    • Lidar + camera
  • Multi-agent SLAM
    • Drone group
    • Multi-user devices
  • Field calibration
    • For life long SLAM
  • More challenge environments
    • Dynamic environments
    • Degraded environment
  • Semantic SLAM
    • Image segmentation and understanding for better feature extraction
    • Semantic point cloud and map for better feature matching
  • Combine with deep learning
    • Learning based feature extraction & matching
    • Learning based relocalization / loop closure detection
    • End-to-end learning based IMU processing