Abstract: 

The domain of outdoor mobile robots opens up many new applications. This includes search and rescue operations, scouting operations, fire fighting, surveillance, etc. Most of these events take place in dangerous environments such as remote areas, places of high radiation, war zones, etc. As such, the deployment of outdoor mobile robots on to the field will not only help save lives but also reduces the danger exposed to search and rescue squads.

This research project aims to develop a vision-based outdoor mobile robot which can autonomously navigate across a large and unstructured environment. The system relies predominantly on visual information and aims to achieve localization by applying an appearance-based approach. In addition, global position will be obtained using an inexpensive GPS receiver to complement the deficiency of visual approaches to operate reliably in low textured or uncluttered environments. Dynamic obstacles will be accounted for by using a real time stereo system and long range depth perception will be achieved by implementing a multibaseline catadioptric stereo system. The surrounding environment will then be reconstructed by taking into account of accuracy and efficiency. Subsequently, obstacles will be probabilistically represented using Occupancy Grid which is then followed by path planning using the infamous Distance Transform. Ultimately, the robot will make well-informed decisions by blending both deliberative and reactive control strategies and probabilistic frameworks to account for the uncertainty in the information provided by the visual sensors and GPS receiver.

About the Speaker: Dennis is a postgraduate research student with the Department of Electrical and Computer Systems Engineering at Monash University