WIT Press


Improving Robot Navigation Using Learning Strategies

Price

Free (open access)

Paper DOI

10.2495/AIENG930341

Volume

1

Pages

7

Published

1993

Size

623 kb

Author(s)

L.M. Gambardella & M. Haex

Abstract

Improving robot navigation using learning strategies L.M. Gambardella & M. Haex Instituto Dalle Molle di Studi suH'Intelligenza CH, Switzerland ABSTRACT While most motion planning techniques do not consider the shape and size of the robot, these characteristics usually influence the performance of the planner. This paper presents a method to improve the performance of grid potential field planners by enhancing them with learning techniques. A neu- ral network technique is proposed to learn the low level motion planning, i.e. chosing the actions to move according to a field thereby avoiding obstacles. INTRODUCTION The purpose of this paper is to describe how learning methods and motion planning techniques can be combined to obtain a performance improvement for motion planning in robotics. Generally, motion planning techniques do not consider characteristics of the shape and size of the robot during the planning phase. The robot is usually represented as a point and its phy

Keywords