Publication detail

The Walking Policies Automatic Generation Using Beam Search

ONDROUŠEK, V. BŘEZINA, T. KREJSA, J.

Czech title

The Walking Policies Automatic Generation Using Beam Search

English title

The Walking Policies Automatic Generation Using Beam Search

Type

conference paper

Language

en

Original abstract

This contribution deals with the problem of automatic generation of different styles of statically stable walking gait of a four-legged robot. The search for appropriate strategies is based on the scanning of state space using an informed search algorithm called A* and an algorithm of beam search. This article compares these two methods and considers the suitability of their use. A tree is used to represent and describe the state space. The tree is generated using a set of elementary production rules chosen in order to obtain an acceptable complexity of the tree and algorithm response. Policies are continuously generated as short-time plans of robot movements. Verification of the discussed algorithms and proposed rules is provided using software simulation.

Czech abstract

This contribution deals with the problem of automatic generation of different styles of statically stable walking gait of a four-legged robot. The search for appropriate strategies is based on the scanning of state space using an informed search algorithm called A* and an algorithm of beam search. This article compares these two methods and considers the suitability of their use. A tree is used to represent and describe the state space. The tree is generated using a set of elementary production rules chosen in order to obtain an acceptable complexity of the tree and algorithm response. Policies are continuously generated as short-time plans of robot movements. Verification of the discussed algorithms and proposed rules is provided using software simulation.

English abstract

This contribution deals with the problem of automatic generation of different styles of statically stable walking gait of a four-legged robot. The search for appropriate strategies is based on the scanning of state space using an informed search algorithm called A* and an algorithm of beam search. This article compares these two methods and considers the suitability of their use. A tree is used to represent and describe the state space. The tree is generated using a set of elementary production rules chosen in order to obtain an acceptable complexity of the tree and algorithm response. Policies are continuously generated as short-time plans of robot movements. Verification of the discussed algorithms and proposed rules is provided using software simulation.

Keywords in English

beam search, walking gait, A star, four-legged

RIV year

2006

Released

31.05.2006

Location

Brno

ISBN

80-214-3195-4

Book

Mendel 2006

Journal

Nezařazené články

Edition number

1

Pages from–to

145–150

Pages count

6

BIBTEX


@inproceedings{BUT20035,
  author="Vít {Ondroušek} and Tomáš {Březina} and Jiří {Krejsa},
  title="The Walking Policies Automatic Generation Using Beam Search",
  journal="Nezařazené články",
  booktitle="Mendel 2006",
  year="2006",
  month="May",
  pages="145--150",
  address="Brno",
  isbn="80-214-3195-4"
}