Sensor-based manipulation planning as a game with nature - Robotics Institute Carnegie Mellon University

Sensor-based manipulation planning as a game with nature

Russell H. Taylor, Matthew T. Mason, and Kenneth Y. Goldberg
Conference Paper, Proceedings of 4th International Symposium on Robotics Research (ISRR '88), pp. 421 - 429, May, 1988

Abstract

This paper explores a game-theoretic approach to automatic planning of sensor-based robotic manipulation programs. To win the game, the robot must provably attain a specified task state. The robot moves by choosing a control signal, and nature moves by choosing a sensor signal. Planning is accomplished by searching the game tree. In some task domains, the approach provides a straightforward method of reasoning about uncertainty, non-deterministic actions, and imperfect sensors. We demonstrate the approach in two different task domains: orienting an object using an instrumented tilting tray: and orienting and grasping an object with an instrumented parallel-jaw gripper.

BibTeX

@conference{Taylor-1988-15690,
author = {Russell H. Taylor and Matthew T. Mason and Kenneth Y. Goldberg},
title = {Sensor-based manipulation planning as a game with nature},
booktitle = {Proceedings of 4th International Symposium on Robotics Research (ISRR '88)},
year = {1988},
month = {May},
editor = {Robert C. Bolles and Bernard Roth},
pages = {421 - 429},
publisher = {MIT Press},
address = {Cambridge MA},
}