AN EFFICIENT APPROACH TO PLANNING IN ASSEMBLY TASKS

• Type: Article in International Scientific Journal
• Date: 1991

Authors

ramos, c (Author) Other; oliveira, e (Author) FEUP

Journal

Title: Lecture Notes in Computer Science

Vol. 5949

Pages: 210-221

ISSN: 0302-9743

Publisher: Springer Nature

Indexing

ISI Web of Knowledge - 1 Citation

Scientific classification

FOS: Natural sciences > Computer and information sciences

Other information

Authenticus ID: P-001-R6R

Abstract (EN): Assembly Robotics is considered a good testbed to evaluate Al planning algorithms. In most cases, these algorithms are tested on simulated and simplified robotic environments. However, the application of such algorithms to real assembly tasks fail since the usual approach, restrictive to the blocks world, is too poor to be generalized for real-life robotic problems. On the other hand, these algorithms are very expensive to be applied on Robotics, where real time problems are often found. The aim of this paper is to describe our on-going work concerning the implementation of an Intelligent and Efficient Planner, the so called High Level Planner (HLP), to deal with real Assembly problems. In our approach some important features such as the conversion of Computer Vision outputs (object positions and orientations) to some inputs used by the Planner (symbolic relationships) as well as the execution of the high level plan by a real robot (RENAULT-APRA) are taken into account. HLP is implemented in PROLOG and is member of a Multi-Agent Community together with other four agents: Object Identifier (VISION), World Descriptor (WD), Models (MODELS) and the Low Level Executor (LLE). The HLP generates efficient plans which are automatically translated to a robot level language by the LLE which directly controls the robot. Two alternative searching methods can be chosen in the plan formulation. The first one uses Best First algorithm, being appropriated for real time constraints dealing. The other algorithm, using a kind of Branch and Bound, needs some more time to be executed, but the best plan is obtained. Both algorithms consider an efficient pruning of incorrect paths of the planning tree.

Language: English

Type (Professor's evaluation): Scientific

No. of pages: 12