A Distributed Human Following Robot using Deep Learning and Sensor Fusion
| dc.contributor.author | Zuñiga, Hiram | |
| dc.contributor.author | Bremer, Luis Bernardo | |
| dc.contributor.author | Orozco Rosas, Ulises | |
| dc.contributor.author | Picos, Kenia | |
| dc.date.accessioned | 2026-11-02T20:22:49Z | |
| dc.date.issued | 2026-11-02 | |
| dc.description.abstract | This work presents a human-following mobile robot using deep learning and a reactive LiDAR-based safety layer to enable robust indoor navigation. Built on the TurtleBot2 platform with a Kobuki base, the system runs a YOLOv3-based model for person detection, a Channel and Spatial Reliability Tracking (CSRT) filter for inter-frame bounding-box propagation, and an HSV histogram identity-locking mechanism to maintain focus on a single target in partially occluded scenes. A 2D RPLiDAR A2M sensor provides an independent reactive safety layer that suppresses velocity commands whenever an obstacle enters the robot’s frontal arc. Computation is partitioned between an embedded NVIDIA Jetson TX2 and an external master laptop. This ROS-distributed architecture design avoids the need for a high-performance computing unit mounted on the robot, such as a workstation or laptop. This supports future scalability and leaves the heavy perception tasks to the ROS Master node. | |
| dc.description.sponsorship | ELSEVIER | |
| dc.description.url | http://www.sciencedirect.com/ | |
| dc.format.page | 2968–2977 | |
| dc.identifier.doi | https://doi.org/10.1016/j.procs.2026.09.372 | |
| dc.identifier.indexacion | Scopus | |
| dc.identifier.uri | https://repositorio.cetys.mx/handle/60000/2106 | |
| dc.language.iso | en_US | |
| dc.relation.ispartofseries | 292 | |
| dc.subject | Mobile robots; deep learning; human tracking; autonomous navigation; sensor fusion | |
| dc.subject.sede | Campus Tijuana | |
| dc.title | A Distributed Human Following Robot using Deep Learning and Sensor Fusion | |
| dc.title.alternative | Procedia Computer Science | |
| dc.type | Article |
