This report details the robust localization and mapping algorithms developed for the Crowdbot project between months M1 and M30. Our proposed solutions are designed with the explicit goal of achieving robot navigation in crowded environments, where many existing methods struggle due to the high degree of dynamic motion around the robot. This report primarily serves as an update on Crowdbot D31 First Release of Localization, Mapping and Local Motion Planning, which covers the work carried out in months M1 to M20. Specifically, we build on Sections 3 and 4 of the previous deliverable and include relevant algorithmic details from D31 of the Crowdbot mapping and localization algorithms to ensure the completeness of the report. As stated in D31, the main challenges that we address are:
- Generating clean and coherent maps of the static environment despite the presence of dynamic obstacles during mapping;
- Achieving fast, and accurate localization when prior information on the robot pose is unavailable.
The main improvements to our localization and mapping strategy from D31 are:
- Explicit removal of LIDAR points that return from pedestrians around the robot, identified using the pedestrian detection algorithms provided by RWTH, prior to map generation, and
- Refinement of our localization precision through algorithmic improvements to our branch and bound search technique that enable maintaining real-time operation.