Purpose This research aims to establish a global positioning system (GPS) waypoint–based autonomous navigation system with vision-assisted object detection to improve UAV performance in open environments. The goal is to achieve correct waypoint tracking, real-time perception of obstacles and safe payload delivery under diverse operational conditions. Design/methodology/approach The unmanned aerial vehicle platform consists of a Radiolink CrossFlight flight controller for stabilization, a Raspberry Pi 4 for computation on board and an Arducam camera for vision processing in real time. YOLOv5 object detection model was used to detect humans, animals and fixed obstacles, while GPS and inertial measurement unit data were combined using a Kalman filter for accurate localization. Structural resilience was confirmed through Finite Element Analysis and open-field test flights were conducted for testing navigation accuracy, payload release and communication delay. Findings Experimental results showed that the designed system maintained an average waypoint-following error of ±15 cm and object detection confidence of more than 90% under normal daylight illumination. The payload delivery mechanism had an accuracy of within 1 m of the target location, and communication latency was less than 100 ms in all test flights. The UAV showed stable control of altitude, effective energy usage and structural integrity under operational stresses, confirming its appropriateness for autonomous outdoor operation. Originality/value This work offers a lightweight, low-cost and modular UAV platform that combines GPS navigation with real-time computer vision for increased situational awareness and autonomy. The system’s scalability and versatility enable it to be used in a wide range of applications including environmental monitoring, surveillance, disaster relief and precision delivery. This paper differs from previous works that depended on heavy sensor fusion or high-end computation by using a lightweight yet powerful solution that can be deployed in the field and scaled up for multi-UAV co-ordination.
Agrawal et al. (2026) studied this question.