Hindawi Publishing Corporation Advances in Artificial Intelligence Volume 2010, Article ID 765876, 20 pages doi:10.1155/2010/765876 Research Article Bootstrap Learning and Visual Processing Management on Mobile Robots Mohan Sridharan Department of Computer Science, Texas Tech University, Lubbock, TX 79409, USA Correspondence should be addressed to Mohan Sridharan, mohan.sridharan@ttu.edu Received 1 October 2009; Accepted 10 November 2009 Academic Editor: Alfons Schuster Copyright © 2010 Mohan Sridharan. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. A central goal of robotics and AI is to enable a team of robots to operate autonomously in the real world and collaborate with humans over an extended period of time. Though developments in sensor technology have resulted in the deployment of robots in specific applications the ability to accurately sense and interact with the environment is still missing. Key challenges to the widespread deployment of robots include the ability to learn models of environmental features based on sensory inputs, bootstrap off of the learned models to detect and adapt to environmental changes, and autonomously tailor the sensory processing to the task at hand. This paper summarizes a comprehensive effort towards such bootstrap learning, adaptation, and processing management using visual input. We describe probabilistic algorithms that enable a mobile robot to autonomously plan its actions to learn models of color distributions and illuminations. The learned models are used to detect and adapt to illumination changes. Furthermore, we describe a probabilistic sequential decision-making approach that autonomously tailors the visual processing to the task at hand. All algorithms are fully implemented and tested on robot platforms in dynamic environments. 1. Introduction An open grand challenge in the field of robotics is to enable widespread deployment of robots in the real world, where they can operate autonomously and collaborate with humans. Addressing this grand challenge would in turn require answers to the following major questions. (i) Autonomous Learning and Adaptation. How to enable a robot to autonomously learn models of environ- mental features based on sensory input, detect envi- ronmental changes, and adapt the learned models in response to such changes? (ii) Processing Management. Given multiple sources of information, which bits of information should be processed, and what processing should be performed in order to achieve a desired goal reliably and efficiently? (iii) Multiagent Coordination. How to enable a team of robots, each with possibly different capabilities and constraints, to collaborate robustly towards a shared objective despite noisy sensing and communication? In this paper, the focus is primarily on developing proba- bilistic methods for Autonomous Learning and Adaptation, and for Processing Management. We propose probabilistic methods that enable a robot to use sensory inputs to learn environmental models and respond to environmental changes. Furthermore, given multiple sources of informa- tion, the robot autonomously tailors the sensory processing to the task at hand. Mobile robots that sense and interact with the environ- ment through a set of sensors and actuators are characterized by the following features and requirements. (i) Features (a) Partial Observability. The true state of the world is not directly observable. The robot can only update its belief, that is, an estimate of the world state by executing actions and observing the noisy outcomes. (b) Nondeterministic Actions and Observations. The outcome of executing actions or making obser- vations based on sensory input is nondeter- ministic, that is, actions and observations are unreliable.