Parallel algorithms for digital map path optimisation
Parallel algorithms for digital map path optimisation
Working out the best route for a land vehicle to take from its current position to a certain destination can be done relatively quickly and easily by a human, given a map of the area. However, if the map is large and complex, and there are many different constraints on the route taken, then it should be faster to find the optimal route using a computer. Moreover, if the process can be distributed to run on several computers in parallel, the results will be obtained even more quickly. This thesis describes an investigation into finding fast and efficient implementations of parallel algorithms to solve the problem, given a digital map of the area. The target architecture is of the distributed memory MIMD variety and for the particular implementation described in this thesis, a transputer array was used. Firstly, a study of sequential optimum path-finding algorithms is given. Out of the four main algorithms: dynamic programming, Dijkstra's algorithm, d'Esopo's algorithm and branch-and-bound, d'Esopo's algorithm is verified by experiment to be the fastest. The results of distributing Yen's algorithm, a version of dynamic programming, and of distributing d'Esopo's algorithm, are then given. Parallel Yen is found to be fast and have an almost linear improvement in performance with the number of transputers used. However, its performance deteriorates rapidly as the map size is increased. Three different parallel versions of d'Esopo's algorithm are described. Although each is found to have high communication overheads, and to be slower than parallel Yen for the relatively small map sizes used, it is shown that the performances of the second and third parallel d'Esopo programs improve relative to Yen's as the map size is increased. Finally, it is shown how one of these parallel d'Esopo algorithms would be suitable for implementing in a distributed architecture vehicle guidance system.
University of Southampton
1992
Hislop, Andrew David
(1992)
Parallel algorithms for digital map path optimisation.
University of Southampton, Doctoral Thesis.
Record type:
Thesis
(Doctoral)
Abstract
Working out the best route for a land vehicle to take from its current position to a certain destination can be done relatively quickly and easily by a human, given a map of the area. However, if the map is large and complex, and there are many different constraints on the route taken, then it should be faster to find the optimal route using a computer. Moreover, if the process can be distributed to run on several computers in parallel, the results will be obtained even more quickly. This thesis describes an investigation into finding fast and efficient implementations of parallel algorithms to solve the problem, given a digital map of the area. The target architecture is of the distributed memory MIMD variety and for the particular implementation described in this thesis, a transputer array was used. Firstly, a study of sequential optimum path-finding algorithms is given. Out of the four main algorithms: dynamic programming, Dijkstra's algorithm, d'Esopo's algorithm and branch-and-bound, d'Esopo's algorithm is verified by experiment to be the fastest. The results of distributing Yen's algorithm, a version of dynamic programming, and of distributing d'Esopo's algorithm, are then given. Parallel Yen is found to be fast and have an almost linear improvement in performance with the number of transputers used. However, its performance deteriorates rapidly as the map size is increased. Three different parallel versions of d'Esopo's algorithm are described. Although each is found to have high communication overheads, and to be slower than parallel Yen for the relatively small map sizes used, it is shown that the performances of the second and third parallel d'Esopo programs improve relative to Yen's as the map size is increased. Finally, it is shown how one of these parallel d'Esopo algorithms would be suitable for implementing in a distributed architecture vehicle guidance system.
This record has no associated files available for download.
More information
Published date: 1992
Identifiers
Local EPrints ID: 461291
URI: http://eprints.soton.ac.uk/id/eprint/461291
PURE UUID: 5ed07879-8d1a-45ae-be23-666d88680a92
Catalogue record
Date deposited: 04 Jul 2022 18:42
Last modified: 04 Jul 2022 18:42
Export record
Contributors
Author:
Andrew David Hislop
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics