Skip to main content
Fachgebiete
Automobil + Motoren Bauwesen + Immobilien Business IT + Informatik Elektrotechnik + Elektronik Energie + Nachhaltigkeit Finance + Banking Management + Führung Marketing + Vertrieb Maschinenbau + Werkstoffe Versicherung + Risiko
DE
EN
Themenseiten
Organisationspsychologie Projektmanagement Marketing Smart Manufacturing
Bücher
Zeitschriften
Weitere Formate
Podcasts Webinare Technik Webinare Wirtschaft Kongresse Veranstaltungskalender Awards
Jetzt Einzelzugang starten
Zugang für Unternehmen
Referenzkunden
ATZ-Fachkongress

Chassis.tech plus 2024


4 Kongresse in einer Veranstaltung

Treffen Sie in München die Fachleute von Chassis, Lenkung, Bremsen sowie Reifen/Räder.
Erweiterte Suche
Anmelden
nach oben

1990 | Buch

Introduction Kapitel lesen Erstes Kapitel lesen

Task-Directed Sensor Fusion and Planning

A Computational Approach

verfasst von: Gregory D. Hager

Verlag: Springer US

Buchreihe : The International Series in Engineering and Computer Science

Enthalten in: Professional Book Archive

Einloggen, um Zugang zu erhalten
insite
SUCHEN

Über dieses Buch

If you have ever hiked up a steep hill to reach a viewpoint, you will know that sensing can involve the expenditure of effort. More generally, the choice of which movement an intelligent system chooses to make is usually based on information gleaned from sensors. But the information required to make the motion decision may not be immediately to hand, so the system . first has to plan a motion whose purpose is to acquire the needed sensor information. Again, this conforms to our everyday experience: I am in the woods and don't know which direction to go, so I climb up to the ridge to get my bearings; I am lost in a new town, so I plan to drive to the next junction where there is sure to be a roadsign, failing that I will ask someone who seems to be from the locality. Why, if experiences such as these are so familiar, has the problem only recently been recognised and studied in Robotics? One reason is that until quite recently Robotics research was dominated by work on robot arms with limited reach and fixed in a workcell.

Inhaltsverzeichnis

Frontmatter
Chapter 1. Introduction
Abstract
A key element of intelligent behavior is the ability to quickly and correctly assess a situation and to act or react accordingly.
Gregory D. Hager
Chapter 2. Modeling Sensors
Abstract
The first step in any fusion problem is to clearly define and model what is being observed, how it is observed, and the accuracy of those measurements. Sensor fusion is, broadly described, the process of using one or more of these sensor models to select a data description that is compatible with observations, and to calculate the accuracy of this description based on the accuracy of the original data. Put in other words, sensor modeling is the process of describing how a sensor images the world, and sensor fusion uses that description to infer what world descriptions are compatible with observed information. Consequently, a reasonable representation for observed structures and an accurate description of the sensor itself are imperative for accurate and efficient sensor fusion.
Gregory D. Hager
Chapter 3. Task Modeling and Decision Making
Abstract
Once we have a properly expressed model that accurately reflects the sensor characteristics, and we have decided what other sensors are observing the same object or structure, the process of sensor fusion is largely reduced to the selection and application of one of a number of well-known methods for parameter estimation to the given problem. Anyone doing sensor fusion should understand what techniques are available, and use the technique that is appropriate for the problem. In Appendix B we briefly describe various methods for parameter estimation, and urge the reader not familiar with the properties of, and relationships between, these techniques to read that section and the given references.
Gregory D. Hager
Chapter 4. Mean-Square Estimation
Abstract
In the decision-theoritic literatur, there are few if any results of the generality required for a substantial portion of the information gathering problems described in the previous two chapters. However, there are well-known results for minimum mean-square error parameter estimation (MMSE) for known linear systems corrupted by Gaussian noise. When there is an accompanying linear dynamic system, the optimal solution to the estimation problem is known as the Kalman filter [Gelb, 1974].1 In the case of nonlinear systems, the most common alternative is to approximate the nonlinear system by a linear one and apply MMSE procedures. These methods have been used in countless applications and are a standard technique in nearly every book on control, decision theory or estimation. The appeal of these methods is their mathematical and computational simplicity, particularly when dealing with dynamic systems—systems whose state evolves over time.
Gregory D. Hager
Chapter 5. Grid-Based Probability Density Methods
Abstract
The generality and computability of the methods described in the Chapter 3 depend largely on the representation of probability distributions functions. A moment representation is attractive since the representation and manipulation of uncertainly is relatively simple, however, as we discussed, it has a number of limitations. Hence, to go further with a Bayesian philosophy we mush find a class of probability distributions that is flexible enough to describe both prior and posterior distributions that is flexible enough to describe both prior and posterior distributions after updating with a nonlinear, coupled sensor description, that can be easily transformed and integrated to accommodate a variety of task descriptions, and that is still computationally tractable.
Gregory D. Hager
Chapter 6. Choosing Viewpoints and Features
Abstract
This chapter is devoted to the problem of choosing sensor control parameters to maximize the performance of the information fusion process for a given task. As described in Chapter 3, we consider the problem of choosing sensor control parameters in terms of the theory of experimental design [Fedorov, 1972; Mendenhall, 1968]. Experimental design is concerned with the problem of maximizing the information gained from an experiment under cost constraints by choosing the most cost-effective experimental parameters. By analogy, our experimental apparatus is the sensing system and associated processing software, our costs are time, sensor use, and information processing resources, and we seek a sensing plan with maximal net payoff to the system.
Gregory D. Hager
Chapter 7. Towards a Task-Level Programming Environment
Abstract
To this point, we have discussed a mathematical framework for task-directed information gathering, described both the mean-square family of estimators and the grid-based method, and discussed their properties. The implementation of mean-square estimation methods is relatively simple and well-known, and a discussion thereof can be found in many of the cited references. The purpose of this chapter is to describe a concrete implementation of the grid-based techniques and to illustrate its application. We feel it is important to discuss this topic as it provides another means of gaining intuition about the flexibility, generality, and practicality of a method.
Gregory D. Hager
Chapter 8. An Experimental System
Abstract
In order to draw together a number of the concepts and ideas in this book, we describe an experimental system and the results of some experiments with the previously described grid-based techniques. This system is not by any manes an extremely complex multi-sensor system, but it serves to test the fusion and information-gathering methods, as well as to illustrate some organizational principles. We will first explain the operation of the basic components and detail a sensor fusion program for the system, and then present the results of several different types of information gathering.
Gregory D. Hager
Chapter 9. Future Extensions
Abstract
We view the work we have discussed to this point as the low-level “kernel” of an intelligent information-gathering system. There are several issues related to information gathering—computational, theoritical, and philosophical—which need to be explored further. The purpose of this chapter is to place this work in a larger context.
Gregory D. Hager
Backmatter
Metadaten
Titel
Task-Directed Sensor Fusion and Planning
verfasst von
Gregory D. Hager
Copyright-Jahr
1990
Verlag
Springer US
Electronic ISBN
978-1-4613-1545-2
Print ISBN
978-1-4612-8828-2
DOI
https://doi.org/10.1007/978-1-4613-1545-2