Edge AI-Enabled Person Detection and LiDAR-Only SLAM for a Low-Cost SAR Robot

Strang, J, Sharp, A and Durieux, O (2026) Edge AI-Enabled Person Detection and LiDAR-Only SLAM for a Low-Cost SAR Robot. In: 2026 10th International Conference on Mechatronics Engineering (ICOM), 8-9 September, Malaysia/Online.

[img]
Preview
Text
WURO_Edge AI.pdf - Accepted Version
Available under License Creative Commons Attribution.

Download (2MB) | Preview

Abstract

Autonomous search and rescue (SAR) robots offer a way to reduce risk to human search teams during disaster first response. However, commercial platforms are too costly for widespread deployment, while low-cost alternatives fail due to wheel odometry errors on debris and insufficient computing power for parallel navigation and person detection. This paper presents a £431.68 prototype built on a Raspberry Pi5, using only commercial off-the-shelf components and opensource software. Two design choices resolve the computational bottleneck. First, the SLAM-Toolbox is configured for odometryless operation, using LiDAR scan matching alone for position estimation. Second, YOLOv8n inference is offloaded to the Sony IMX500 on-sensor Digital Signal Processor, freeing the CPU for simultaneous SLAM and Nav2 navigation. A Foxglove operator dashboard integrates the live map, detection camera, and battery telemetry into a single interface. Over ten indoor test runs, the system achieved a 100% success rate in six consecutive runs after parameter tuning (runs 4–9), with an average Time to Detection of 2–3 minutes in a 55 m2 three-room course.

Item Type: Conference or Workshop Item (Paper)
Divisions: Applied Science, Computing and Engineering
Depositing User: Emma Harrison
Date Deposited: 11 Nov 2026 16:01
Last Modified: 11 Nov 2026 16:01
URI: https://wrexham.repository.guildhe.ac.uk/id/eprint/18495

Actions (login required)

Edit Item Edit Item