Learning Temporal Co-Observability Relationships for Lifelong Robotic Mapping
نویسندگان
چکیده
This paper reports on a method that learns the temporal co-observability relationships for exemplar views of a dynamic environment collected during long-term robotic mapping. These relationships are efficiently captured using a Chow-Liu tree approximation and allow one to predict which exemplars will be observed by the robot given the robot’s recent observations. For example, these learned relationships can encode scene dependent changes in lighting due to time of day and weather, without explicitly modeling them. Preliminary experimental results are shown using images from 17 fixed locations collected hourly over the course of 116 days.
منابع مشابه
Analysis of Memory-Based Learning Schemes for Robot Navigation in Discrete Grid-Worlds with Partial Observability
Abstract In this paper we tackle the problem of robot navigation in discrete grid-worlds using memory-based learning schemes. Different memory-based approaches are tested for navigating an agent across a discrete but partially observable world, and the significance of memory structure is examined. Further, the effects of additional memory hierarchies and multi-level learning frameworks are anal...
متن کاملRobot Navigation in Partially Observable Domains using Hierarchical Memory-Based Reinforcement Learning
In this paper, we attempt to find a solution to the problem of robot navigation in a domain with partial observability. The domain is a grid-world with intersecting corridors, where the agent learns an optimal policy for navigation by making use of a hierarchical memory-based learning algorithm. We define a hierarchy of levels over which the agent abstracts the learning process, as well as it...
متن کاملSCHOOLING, LIFELONG LEARNING, AND THE FUTURE Selected Insights from OECD Analyses
This paper draws on a selection of recent OECD work to explore the relationships between schooling and lifelong learning and relate these to the Schooling for Tomorrow scenarios. It is an initial version of one of the transversal analyses contributing to the analytical component of this project, and provides background to the Forum session on Scenario Analysis and Indicators.
متن کاملNaive Augmenting Q-Learning to Process Feature-Based Representations of States
Temporal difference algorithms perform well on discrete and small problems. This paper proposes a modification of the Q-learning algorithm towards natural ability to receive a feature list instead of an already identified state in the input. Complete observability is still assumed. The algorithm, Naive Augmenting Q-Learning, has been designed through building a hierarchical structure of input f...
متن کاملAn Active Stereo Vision-Based Learning Approach for Robotic Tracking, Fixating and Grasping Control
Vision-based robotic tracking, fixating and grasping control depends on many environmental factors in an unknown environment. The robot control systems lack robustness, and the calibration of the CCD cameras is very slow and tedious in the existing methods. Although the binocular cameras can solve some of these problems, it is necessary to rely on the time consuming and complicated 3-D reconstr...
متن کامل