Learning Situation-Dependent Costs: Using Execution to Refine Planning Models

Physical environments are so complex that it is hard to hand-tune all of the domain knowledge, especially to model the dynamics of the environment. The work presented in this paper explores machine learning techniques to autonomously identify situations in the environment that a ect plan quality. We introduce the concept of situation-dependent costs, where situational features can be attached to the costs used by the path planner. These costs e ectively diagnose and predict situations the robot encounters so that the planner can generate paths that are appropriate for each situation. We present an implementation of our situationdependent learning approach in a real robotic system, Rogue. Rogue learns situation-dependent costs for arcs in a topological map of the environment; these costs are then used by the path planner to predict and avoid failures. In this article, we present the representation of the path planner and the navigationmodules, and describe the execution trace. We show how training data is extracted from the execution trace. We present experimental results from a simulated, controlled environment as well as from data collected from the actual robot. Our approach e ectively re nes models of dynamic systems and improves the efciency of generated plans.

[1]  L. Rabiner,et al.  An introduction to hidden Markov models , 1986, IEEE ASSP Magazine.

[2]  Reid G. Simmons,et al.  Passive Distance Learning for Robot Navigation , 1996, ICML.

[3]  K. Haigh,et al.  A Layered Architecture for Ooce Delivery Robots , 1997 .

[4]  Thomas G. Dietterich.  What is machine learning? , 2020, Archives of Disease in Childhood.

[5]  David Kortenkamp,et al.  Topological Mapping for Mobile Robots Using a Combination of Sonar and Vision Sensing , 1994, AAAI.

[6]  Andrew W. Moore,et al.  Reinforcement Learning: A Survey , 1996, J. Artif. Intell. Res..

[7]  Richard A. Becker,et al.  The New S Language , 1989 .

[8]  Richard Goodwin.  Meta-Level Control for Decision-Theoretic Planners , 1996 .

[9]  R. R. Murphy,et al.  Learning the expected utility of sensors and algorithms , 1994, Proceedings of 1994 IEEE International Conference on MFI '94. Multisensor Fusion and Integration for Intelligent Systems.

[10]  Douglas J. Pearson.  Learning Procedural Planning Knowledge in Complex Environments , 1996, AAAI/IAAI, Vol. 2.

[11]  Katharina Morik,et al.  Learning Concepts from Sensor Data of a Mobile Robot , 2005, Machine Learning.

[12]  Karen Zita Haigh,et al.  A Layered Architecture for O ce Delivery Robots , 1997 .

[13]  Wei-Min Shen,et al.  Autonomous learning from the environment , 1994 .

[14]  Cristina Baroglio,et al.  Learning Controllers for Industrial Robots , 2005, Machine Learning.

[15]  Trevor Hastie,et al.  Statistical Models in S , 1991 .

[16]  Sebastian Thrun,et al.  A Bayesian Approach to Landmark Discovery and Active Perception in Mobile Robot Navigation , 1999 .

[17]  Ming Tan,et al.  Cost-sensitive robot learning , 1991 .

[18]  Dean A. Pomerleau,et al.  Neural Network Perception for Mobile Robot Guidance , 1993 .

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