As robots become easier to develop and cheaper to field, multi-robot systems are becoming more commonplace. This project explores the problem of mapping an environment using the combination of aerial robots and ground-based robots. The problem is to accurately match the sensor readings between these different modalities. An object seen from the ground can look significantly different when it is observed from the air. Additionally, GPS information may be unavailable to some or all of the robots, due to occlusions by trees or buildings or jamming. Our approach is to build up local maps from the sensors in each modality using SLAM techniques and then match features extracted from the maps in a Bayesian filter framework. The challenge is to find features that can be accurately and reliably matched between the aerial and ground-based maps.

Sponsor:

Korean Agency for Defense Development