1. FUTURE SLAM (신동원, GIST)

연구 동향:

정확한 3D 재구성 및 장면 이해
- MID-Fusion (Octotree-based Object-Level Multi-Instance Dynamic SLAM)
  - Semantic and instance segmentation
  - Robust in a dynamic environment
- Semantic KITTI Dataset (A dataset for semantic segmentation of point cloud sequences)
  - Lidar point cloud with semantic labelling
  - For development of the scenario-oriented localization and mapping
디바이스의 경량화, 소형화
- Nvidia Jetson Nano, Google Coral, Raspberry Pi, Arduino
Multi-agent, Collaborative-SLAM
- Maplab (An Open Framework for research in visual-inertial mapping and localisation)
  - co-registered maps from multiple sessions together
  - Cloud system ??의 사용 가능성을 봄
- Bosch - Bosch Localisation for automated driving
- Lyft - Layerd map
  - Map data hierarchy (피라미드 꼭대기부터: Real-time layer, Map priors layer, semantic map layer, geometric map layer, base map layer (SD map))
- Naver Labs
  - Self-updating map (CVPR 2019)
Spatial AI로서의 SLAM
- FutureMapping : The Computational Structure of Spatial AI Systems