Rapid Concept Learning for Mobile Robots

Authors:
Sridhar Mahadevan;Georgios Theocharous;Nikfar Khaleeli
Affiliations:
Department of Computer Science, Michigan State University, East Lansing, MI 48824. E-mail: mahadeva@cps.msu.edu, theochar@cps.msu.edu;Department of Computer Science, Michigan State University, East Lansing, MI 48824. E-mail: mahadeva@cps.msu.edu, theochar@cps.msu.edu;Wind River Systems, Alameda, CA 94501. E-mail: nikfar@wrs.com
Venue:
Machine Learning - Special issue on learning in autonomous robots
Year:
1998

Citing 13
Cited 3

Fundamentals of digital image processing

Fundamentals of digital image processing
Robotics in service

Robotics in service
Parallel distributed processing: explorations in the microstructure of cognition, vol. 1: foundations

Parallel distributed processing: explorations in the microstructure of cognition, vol. 1: foundations
Automatic programming of behavior-based robots using reinforcement learning

Artificial Intelligence
Cost-Sensitive Learning of Classification Knowledge and Its Applications in Robotics

Machine Learning
Active learning for vision-based robot grasping

Machine Learning - Special issue on robot learning
Xavier: a robot navigation architecture based on partially observable Markov decision process models

Artificial intelligence and mobile robots
Reinforcement Learning

Reinforcement Learning
Machine Learning

Machine Learning
Navigating Mobile Robots: Systems and Techniques

Navigating Mobile Robots: Systems and Techniques
Robot Learning

Robot Learning
Induction of Decision Trees

Machine Learning
Learning One More Thing

Learning One More Thing

Continuous categories for a mobile robot

AAAI '99/IAAI '99 Proceedings of the sixteenth national conference on Artificial intelligence and the eleventh Innovative applications of artificial intelligence conference innovative applications of artificial intelligence
Symbolic Place Recognition in Voronoi-Based Maps by Using Hidden Markov Models

Journal of Intelligent and Robotic Systems
Object learning and detection using evolutionary deformable models for mobile robot navigation

Robotica

Quantified Score

Hi-index	0.00

Visualization

Abstract

Concept learning in robotics is an extremely challenging problem:sensory data is often high-dimensional, and noisy due to specularities andother irregularities. In this paper, we investigate two general strategiesto speed up learning, based on spatial decomposition of the sensoryrepresentation, and simultaneous learning of multiple classes using a sharedstructure. We study two concept learning scenarios: a hallway navigationproblem, where the robot has to induce features such as“opening” or “wall”. The second task is recycling,where the robot has to learn to recognize objects, such as a “trashcan”. We use a common underlying function approximator in both studiesin the form of a feedforward neural network, with several hundred inputunits and multiple output units. Despite the high degree of freedom affordedby such an approximator, we show the two strategies provide sufficient biasto achieve rapid learning. We provide detailed experimental studies on anactual mobile robot called PAVLOV to illustrate the effectiveness of thisapproach.