Collaborative SLAM
SLAM (simultaneous localization and mapping) builds a map and, at the same time, finds the position of a vehicle in that map.
In conventional SLAM, the map often belongs to one vehicle only. The memory of that vehicle limits the size of the map, and the vehicle can only navigate to places that it has already seen.
Cartio Robotics develops a collaborative solution. A fleet of autonomous vehicles builds the map together, and the software distributes the map and other information to every vehicle. The vehicles can share information over any communication medium that is available: 5G, Starlink or mesh-based vehicle-to-vehicle communication. The two videos below show how it works.
Four agents, one map
The first video shows four agents that travel in the same area. At the start, all four are at the origin, and they do not know their positions relative to each other yet. As the agents explore, they find common features in each other’s maps. Each match lets the software update their relative positions. The result is one accurate map of the area, larger than the part that any one agent has seen.
Map sharing over mesh radio
The second video shows one of these communication media: mesh-based vehicle-to-vehicle communication. It is an illustration with three vehicles. Each vehicle builds its map from lidar, camera, IMU and GNSS data. GNSS is satellite positioning, such as GPS.
The vehicles share their maps over mesh-based tactical radio. In a mesh network, each radio passes data on to the other radios, with no central station. When the radios find a communication link, the vehicles share the most valuable information first. As a result, all vehicles get a common situational awareness: the same picture of the area.