AI Enabled Robot for Data Collection in Unreachable and Extreme Environment

Abstract

This article outlines a groundbreaking approach
to gathering data in hazardous or inaccessible
environments through the utilization of innovative
robotics. These robots are specifically designed to navigate
and collect vital information from areas too dangerous or
remote for human exploration, enabling unprecedented
research opportunities. Central to this advancement is the
integration of artificial intelligence (AI) support within
drones, endowed with human recognition capabilities . By
analyzing live drone footage using advanced pattern
recognition techniques like YOLO (You Only Look Once),
these drones achieve high-precision, real-time human
detection. Equipped with an array of sensors, including
cameras and GPS tracking systems, these autonomous
robots are poised to revolutionize data collection and
analysis in challenging environments. The proposed drone
system represents a stateof-the-art solution to object
detection challenges in harsh settings. By amalgamating
cutting-edge technologies such as GPS tracking, obstacle
avoidance, altitude holding features, and the YOLOv8
algorithm, this system offers unparalleled real-time
monitoring and situational awareness capabilities.
Leveraging GPS monitoring for efficient object localization
and the YOLOv8 algorithm for quick and accurate
detection, coupled with the drone’s adeptness at navigating
difficult terrain and maintaining stable flight, ensures
consistent and dependable video feed quality. Moreover, a
comprehensive strategy is employed to enhance safety by
mitigating potential hazards while simultaneously boosting
operational efficiency. This drone system holds promise for
the delivering of the exceptional performance and
invaluable insights in the face of challenging
circumstances, whether deployed for environmental
monitoring, surveillance missions, or search and rescue
operations. The methodology for object detection using
YOLOv8 involves a series of steps including pre-processing
the input video, running the object detection model,
initializing object post-processing, detecting objects over
the frame, periodically re-detecting objects, and visualizing
the results. Testing was conducted using the COCO
dataset, which encompasses various lighting conditions,
with datasets divided into testing, validation, and training
categories to ensure robust performance evaluation. Photos
with a resolution of 640 × 640 were utilized for
experimentation, underscoring the efficacy of the
proposed approach in addressing object detection
challenges across diverse environmental conditions.