Unsupervised learning of probabilistic models for robot navigation

Navigation methods for office delivery robots need to take various sources of uncertainty into account in order to get robust performance. In previous work, we developed a reliable navigation technique that uses partially observable Markov models to represent metric, actuator and sensor uncertainties. This paper describes an algorithm that adjusts the probabilities of the initial Markov model by passively observing the robot's interactions with its environment. The learned probabilities more accurately reflect the actual uncertainties in the environment, which ultimately leads to improved navigation performance. The algorithm, an extension of the Baum-Welch algorithm, learns without a teacher and addresses the issues of limited memory and the cost of collecting training data. Empirical results show that the algorithm learns good Markov models with a small amount of training data.

[1]  Drew McDermott,et al.  Error correction in mobile robot map learning , 1992, Proceedings 1992 IEEE International Conference on Robotics and Automation.

[2]  Leslie Pack Kaelbling,et al.  Inferring finite automata with stochastic output functions and an application to map learning , 1992, 26th Annual Symposium on Foundations of Computer Science (sfcs 1985).

[3]  Alexander Zelinsky,et al.  A mobile robot exploration algorithm , 1992, IEEE Trans. Robotics Autom..

[4]  S. Sitharama Iyengar,et al.  Robot navigation algorithms using learned spatial graphs , 1986, Robotica.

[5]  Pierre A. Devijver,et al.  Baum's forward-backward algorithm revisited , 1985, Pattern Recognit. Lett..

[6]  Sebastian Thrun,et al.  Exploration and model building in mobile robot domains , 1993, IEEE International Conference on Neural Networks.

[7]  S. Sitharama Iyengar,et al.  Robot navigation in unknown terrains using learned visibility graphs. Part I: The disjoint convex obstacle case , 1987, IEEE Journal on Robotics and Automation.

[8]  Maja J. Mataric,et al.  Environment learning using a distributed representation , 1990, Proceedings., IEEE International Conference on Robotics and Automation.

[9]  Illah R. Nourbakhsh,et al.  DERVISH - An Office-Navigating Robot , 1995, AI Mag..

[10]  Benjamin Kuipers,et al.  A Robust, Qualitative Method for Robot Spatial Learning , 1988, AAAI.

[11]  Reid G. Simmons,et al.  Probabilistic Robot Navigation in Partially Observable Environments , 1995, IJCAI.

[12]  Vladimir J. Lumelsky,et al.  Dynamic path planning in sensor-based terrain acquisition , 1990, IEEE Trans. Robotics Autom..

[13]  S. Sitharama Iyengar,et al.  A 'retraction' method for learned navigation in unknown terrains for a circular robot , 1991, IEEE Trans. Robotics Autom..