Description

The main objective of this thesis is to contribute to the state of the art in 3D outdoor navigation by developing an accurate exploration strategy using a 3D range sensing as the main sensor for terrain perception. Specifically, we work in pedestrian urban areas for a robotic guidance and surveillance application.

Expected contributions

• An on-line robust range data registration
– To provide a rich enviroment representation capable of representing accurately the robot’s surroundings at low computational cost and with reduced effects from the 3D sensor noise.
– Use a new hierarchical nearest neighbor search, and filtering strategies to get a better registration.
– Provide a state of the art data fusion and inference mechanism to refine pose estimates from the registred 3D data with first order uncertainties.
– Computing first order uncertainty propagation of ICP metrics.

• Novel exploration techniques using 3D range data.
– To develop a compact and consistent representation extracting higher order entities/symbols from the registered raw data maps.
– To study the advantage of using multi-resolution based maps to reduce the space state for exploration.
– To propose a finite-horizon strategy for exploration.

The work is under the scope of the following projects:

• URUS: Ubiquitous networking robotics in urban settings (web)
• PAU: Percepción y acción ante incertidumbre (web)
• EARMU: Exploración activa para robots móviles urbanos (web)
• PAU+: Perception and Action in Robotics Problems with Large State Spaces (web)