Understanding Intentions
My work to date has focused on the intent recognition problem: given a video of agents (human or robot) performing various activities, determine for each agent the activity that it is performing, along with the intentional (goal-directed) mental states of each agent. These mental states can then be used by the observer for the purposes of control or prediction.
Papers
- R. Kelley, C. King, A. Tavakkoli, M. Nicolescu, M. Nicolescu, G. Bebis, "An Architecture for Understanding Intent Using a Novel Hidden Markov Formulation", International Journal of Humanoid Robotics, Special Issue on Cognitive Humanoid Robots, 2008.
- A. Tavakkoli, R. Kelley, C. King, M. Nicolescu, M. Nicolescu, G. Bebis, "A Vision-Based Architecture for Intent Recognition", Proceedings of the International Symposium on Visual Computing, Vol. II, pp 173-182, 2007.
- R. Kelley, A. Tavakkoli, C. King, M. Nicolescu, M. Nicolescu, G. Bebis, "Understanding Human Intentions via Hidden Markov Models in Autonomous Mobile Robots", Proceedings of the 3rd International Conference on Human-Robot Interaction, 2008.
- A. Tavakkoli, R. Kelley, C. King, M. Nicolescu, M. Nicolescu, G. Bebis, "A Visual Tracking Framework for Intent Recognition in Videos", Proceedings of the International Symposium on Visual Computing, 2008.
Genetic Algorithms
After spending some time studying genetic algorithms, I took an interest in applying them to problems involving statistic inference. Since I was working with Hidden Markov Models at the time, it seemed natural to play with the idea of GAs for solving the Decoding problem for HMMs. Of course, the Viterbi algorithm gives an optimal solution for this problem, but I argue in the paper below that a GA performs competitively with Viterbi, and has attractive characteristics such as conceptual simplicity and easy parallelizability.
Papers
- R. Kelley, M. Nicolescu, M. Nicolescu, S. Louis, "An Evolutionary Approach to Maximum Likelihood Estimation for Generative Stochastic Models", Proceedings of the 40th International Symposium on Robotics, 2009.
Linguistics and Robotics
I've just started looking into the relationship between linguistic structures and the kinds of structures that are typical in robotics (particularly behaviors). The hope is that ideas from linguistics can be used to guide practical and theoretical advances in robotics and human-robot interaction. The extended abstract listed below suggests one way in which syntactic structures can be used to organize robot behaviors.
Papers
- R. Kelley, M. Nicolescu, M. Nicolescu, "Grammar-Based Robot Control", Proceedings of the 8th International Conference on Autonomous Agents and Multiagent Systems, 2009.
Roboethics and Legal Issues in Robotics
As robots become ubiquitous in everyday life, it is inevitable that disputes involving the use and misuse of robots will become more frequent as well. My work in this area is aimed at setting down some of the foundations that will allow those disputes to be effectively adjudicated. This is joint work with Enrique Schaerer.
Papers
- E. Schaerer, R. Kelley, M. Nicolescu, "Robots as Animals: A Framework for Liability and Responsibility in Human-Robot Interactions", Proceedings of the 18th IEEE Symposium on Robot and Human Interactive Communication, 2009.