Artificial intelligence and machine learning for efficient minefield clearance

Bruckbauer, Alexander and Grout, Vic (2020) Artificial intelligence and machine learning for efficient minefield clearance. [Report]

[img]
Preview
Text
GURO_REP_439_Alexander Bruckbauer and Vic Grout article - CIEDR SS2020.pdf - Published Version

Download (1MB) | Preview

Abstract

Landmines, particularly anti-personnel mines, are dreadful instruments of war. Mines can remain in the ground for decades and injure or kill long after the original conflict. Clearing mines is a dangerous, time consuming and expensive task. Fortunately, mine clearing already has well established and documented processes. To further support these efforts a new research project has started at Wrexham Glyndŵr University to explore the use of machine learning to create a prediction model able to better suggest the positions of hidden landmines based on locations of those already found. Research in psychology and computer science demonstrates the difficulty for humans and machines to create true randomness in their actions. The project will investigate whether it is possible to discover hidden patterns or sequences in mine deployment that could give hints where to look for more. The advantage of the envisioned technique is a lightweight data set only comprising numerical values and their simple acquisition in the field. The proposed system will support – not replace – conventional technology. Although machine learning and A.I. can discover structures, patterns and sequences in a huge data set, that humans cannot, it remains a form of prediction. The aim is therefore not to declare the ground safe (‘cleared’) but to give suggestions where additional explosives are likely to be found and thus, it is proposed, help to direct mine clearing resources better.

Item Type: Report
Divisions: Applied Science, Computing and Engineering
Depositing User: Hayley Dennis
Date Deposited: 03 Jun 2020 13:26
Last Modified: 03 Jun 2020 13:26
URI: https://glyndwr.repository.guildhe.ac.uk/id/eprint/17598

Actions (login required)

Edit Item Edit Item