Parallel distributed processing: explorations in the microstructure of cognition, vol. 2: psychological and biological models
Fundamentals of digital image processing
Fundamentals of digital image processing
Robotics in service
Automatic programming of behavior-based robots using reinforcement learning
Artificial Intelligence
Active learning for vision-based robot grasping
Machine Learning - Special issue on robot learning
Learning concepts from sensor data of a mobile robot
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
Machine Learning
Navigating Mobile Robots: Systems and Techniques
Navigating Mobile Robots: Systems and Techniques
Introduction to Reinforcement Learning
Introduction to Reinforcement Learning
Robot Learning
Machine Learning
Acting Uncertainty: Discrete Bayesian Models for Mobile-Robot Navigation
Acting Uncertainty: Discrete Bayesian Models for Mobile-Robot Navigation
Learning One More Thing
Mobile Robot Learning by Self-Observation
Autonomous Robots
Hi-index | 0.00 |
Concept learning in robotics is an extremely challengingproblem: sensory data is often high dimensional, and noisy due tospecularities and other irregularities. In this paper, we investigatetwo general strategies to speed up learning, based on spatialdecomposition of the sensory representation, and simultaneous learningof multiple classes using a shared structure. We study two conceptlearning scenarios: a hallway navigation problem, where the robot hasto induce features such as “opening” or “wall”. The second task isrecycling, where the robot has to learn to recognize objects, such asa “trash can”. We use a common underlying function approximator inboth studies in the form of a feedforward neural network, with severalhundred input units and multiple output units. Despite the highdegree of freedom afforded by such an approximator, we show the twostrategies provide sufficient bias to achieve rapid learning. Weprovide detailed experimental studies on an actual mobile robot calledPAVLOV to illustrate the effectiveness of this approach.