by L Mösenlechner, A Müller and M Beetz
Abstract:
We investigate the plan-based control of physically and sensorically realistic simulated autonomous mobile robots performing everyday pick-and-place tasks in human environments, such as table setting. Our approach applies AI planning techniques to transform default plans that can be inferred from instructions for activities of daily life into flexible, high-performance robot plans. To find high performance plans the planning system applies transformations such as carrying plates to the table by stacking them or leaving cabinet doors open while setting the table, which require substantial changes of the control structure of the intended activities. We argue and demonstrate that applying AI planning techniques directly to concurrent reactive plan languages, instead of using layered software architectures with different languages, enables the robot action planner to achieve substantial performance improvements (23% - 45% depending on the tasks). We also argue that the transformation of concurrent reactive plans is necessary to obtain the results. Our claims are supported by extensive empirical investigations in realistic simulations.
Reference:
High Performance Execution of Everyday Pick-and-Place Tasks by Integrating Transformation Planning and Reactive Execution (L Mösenlechner, A Müller and M Beetz), In Proceedings of the 1st International Workshop on Cognition for Technical Systems, München, Germany, 6-8 October, 2008.
Bibtex Entry:
@inproceedings{mosenlechner_high_2008,
author = {L Mösenlechner and A Müller and M Beetz},
title = {High Performance Execution of Everyday Pick-and-Place Tasks by Integrating
Transformation Planning and Reactive Execution},
booktitle = {Proceedings of the 1st International Workshop on Cognition for Technical
Systems, München, Germany, 6-8 October},
year = {2008},
abstract = {We investigate the plan-based control of physically and sensorically
realistic simulated autonomous mobile robots performing everyday
pick-and-place tasks in human environments, such as table setting.
Our approach applies {AI} planning techniques to transform default
plans that can be inferred from instructions for activities of daily
life into flexible, high-performance robot plans. To find high performance
plans the planning system applies transformations such as carrying
plates to the table by stacking them or leaving cabinet doors open
while setting the table, which require substantial changes of the
control structure of the intended activities. We argue and demonstrate
that applying {AI} planning techniques directly to concurrent reactive
plan languages, instead of using layered software architectures with
different languages, enables the robot action planner to achieve
substantial performance improvements (23\% - 45\% depending on the
tasks). We also argue that the transformation of concurrent reactive
plans is necessary to obtain the results. Our claims are supported
by extensive empirical investigations in realistic simulations.},
}