by M Beetz, T Arbuckle, A Cremers and M Mann
Abstract:
We present the design of a programming system for IP routines which satisfies the requirements above. Our solution consists of RECIPE, a dynamically loadable, modular architecture in a distributed robot control system that provides the basic IP functionality and manages images and other IP data structures. It provides a variety of standard IP routines such as edge detectors, convolutions, noise reduction, segmentation, etc. RPLIP, an extension of the abstract machine provided by the robot control/plan language RPL. RPLIP provides suitable abstractions for images, regions of interest, etc, and supports a tight integration of the vision routines into the robot control system. Image Processing Plans that provide various methods for combining IP methods into IP pipelines. IP plans support the implementation of robust vision routines and the integration of other sensors such as laser range finders and sonars for object recognition tasks and scene analysis. Since vision routines are RPL programs, they can be constructed, revised, and reasoned about while the robot control program is being executed.
Reference:
Transparent, Flexible, and Resource-adaptive Image Processing for Autonomous Service Robots (M Beetz, T Arbuckle, A Cremers and M Mann), In Procs. of the 13th European Conference on Artificial Intelligence (ECAI-98) (H. Prade, ed.), 1998.
Bibtex Entry:
@inproceedings{beetz_transparent_1998,
author = {M Beetz and T Arbuckle and A Cremers and M Mann},
title = {Transparent, Flexible, and Resource-adaptive Image Processing for
Autonomous Service Robots},
booktitle = {Procs. of the 13th European Conference on Artificial Intelligence
({ECAI-98)}},
year = {1998},
editor = {Prade, H.},
pages = {632–636},
abstract = {We present the design of a programming system for {IP} routines which
satisfies the requirements above. Our solution consists of {RECIPE},
a dynamically loadable, modular architecture in a distributed robot
control system that provides the basic {IP} functionality and manages
images and other {IP} data structures. It provides a variety of standard
{IP} routines such as edge detectors, convolutions, noise reduction,
segmentation, etc. {RPLIP}, an extension of the abstract machine
provided by the robot control/plan language {RPL.} {RPLIP} provides
suitable abstractions for images, regions of interest, etc, and supports
a tight integration of the vision routines into the robot control
system. Image Processing Plans that provide various methods for combining
{IP} methods into {IP} pipelines. {IP} plans support the implementation
of robust vision routines and the integration of other sensors such
as laser range finders and sonars for object recognition tasks and
scene analysis. Since vision routines are {RPL} programs, they can
be constructed, revised, and reasoned about while the robot control
program is being executed.},
}