Katrina Samperi

My current interests focus around mapping for virtual mobile agents. The problem I am specifically interested in is how to enable the agent to explore and generate a map of large scale, dynamic and unstructured spaces.

Current research tends to focus on small scale, static environments as the focus of what they are doing is not on the mapping itself, but is on the process of exploration and navigation. By using virtual agents two aspects of the problem become simplified, how does the agent sense the environment and how does it know its current location. By doing this I can focus on how to represent a map of this environment and how to generate it successfully.

Improving probabilistic roadmaps using trails
December 2011.
This presentation is on some recent work done on improving the generation of probabilistic roadmaps using trail information. Trails have been used in architecture (desire lines) to show where people walk in the environment. They have also been used to aid human navigation in virtual worlds. We observe these trails and use them to influence the generation of a probabilistic road map. We find that when trails and random approaches are combined we get the best results, giving maps with the highest success rate for route planning and the shortest routes.

IR Lab presentation
The idea behind this presentation was to give some of my ideas and get feedback from a group of people about how to structure this, and where to progress from this point

A tale of two cities - Navigation in virtual worlds.
January 2009
This presentation was prepared for the Research Skills module. It is a brief overview of some of the things that initially interested me.

Using Trails to aid Virtual Agent Navigation
This poster was initially created for the graduate school poster conference. It was also presented at the British Science Festival in 2010

Thesis Proposal
Mapping in large scale, dynamic virtual environments

RSMG4

RSMG5