Virtual Clones: Data-Driven Social Navigation
Millions of participants inhabit online virtual worlds such as SL and engage in a wide range of activities, some of which require the performance of tedious tasks. Our goal is to develop a virtual proxy that would replace human participants in online virtual worlds. The proxy should be able to perform simple tasks on behalf of its owner, similar to the way the owner would have performed it. In this paper we focus on the challenge of social navigation. We use a data-driven approach based on recording human participants in the virtual environment; this training set, with a machine learning approach, is then used to control an agent in real time who is performing the same social task. We evaluate our method based on data collected from different participants.
KeywordsNavigation behavioral cloning imitation learning proxy data-driven approach
Unable to display preview. Download preview PDF.
- 1.Boellstorff, T.: Coming of Age in Second Life. Princeton University Press, Princeton (2008)Google Scholar
- 2.Friedman, D., Steed, A., Slater, M.: Spatial Social Behavior in Second Life. In: Proc. Intelligent Virtual Agent, pp. 252–263 (2007)Google Scholar
- 4.Anderson, C.W., Draper, B.A., Peterson, D.A.: Behavioral Cloning of Student Pilots with Modular Neural Networks. In: Proceedings of the Seventeenth International Conference on Machine Learning, pp. 25–32 (2007)Google Scholar
- 5.Isaac, A., Sammut, C.: Goal-directed Learning to Fly. In: ICML, pp. 258–265 (2003)Google Scholar
- 6.Morales, E.F., Sammut, C.: Learning to fly by combining reinforcement learning with behavioural cloning. In: Proceedings of the Twenty-first International Conference on Machine Learning, Banff, Alberta, Canada, pp. 76–83 (2004)Google Scholar
- 7.Hall, E.T.: The Hidden Dimension. Doubleday, New York (1966)Google Scholar
- 8.Hall, E.T.: The Silent Language. Doubleday, New York (1959)Google Scholar