The broad aim of this project is to implement and test visual SLAM (Simultaneous Localization and Mapping) outdoors and in a multi-agent
domain. The first goal is to compare the performance of VSLAM outdoors and indoors (for which the software was originally designed). Once a baseline
set of experiments are completed, then we will investigate the possibility of improving localization and mapping when multiple agents are involved and
landmarks are shared. The software being used is Evolution's ERSP 3.0, of which
VSLAM is a part.
Specifically, we are interested in
- Comparing VSLAM performance indoors and outdoors
- Designing and evaluating modifications to VSLAM to make it more suited for outdoors
- Investigating the communication of landmarks among agents to improve localization and mapping, both in terms of speed and accuracy.
|