Cooperative swarm localization and mapping with inter-agent ranging

Compared to a single robot, a swarm system can conduct a given task in a shorter time, and it is more robust to system failures of each agent. To successfully execute cooperative missions with multiple agents, accurate relative positioning is important. If global positioning (e.g. with a GNSS-based positioning) is available, we can easily compute relative positions. In environments where a global positioning system is unreliable or unavailable, visual odometry can be applied for estimating each agent’s egomotion, by exploiting onboard cameras. Using these self-localization results, relative positions between agents can be estimated, once the relative geometry between agents is initialized. However, since visual odometry is a dead-reckoning process, the estimation errors accumulate inherently without bounds. We propose a cooperative localization method using visual odometry and inter-agent range measurements. Using the proposed method, we can reduce the drifts in position estimates with very modest requirements on the communication channel between agents.      «

Compared to a single robot, a swarm system can conduct a given task in a shorter time, and it is more robust to system failures of each agent. To successfully execute cooperative missions with multiple agents, accurate relative positioning is important. If global positioning (e.g. with a GNSS-based positioning) is available, we can easily compute relative positions. In environments where a global positioning system is unreliable or unavailable, visual odometry can be applied for estimating each...     »