Low-Cost Modular Design for Autonomous Multi-Task Drone System in Complex GPS-Denied Environments
Autonomous drone system for GPS-denied environments. Three-layer architecture integrating visual navigation and PID control, 83% task success rate, landing accuracy better than 10cm.
Development and Integration of Autonomous Multi-Task Drone System
Achieving Complex Autonomous Functions Under Resource Constraints
This project documents the complete development journey of a low-cost, modular autonomous drone system designed to execute multiple tasks in complex GPS-denied environments. This is a true system integration challenge—achieving complex autonomous functions through innovative modular design and refined control algorithms under strict hardware and cost constraints.
The system is built on a centralized three-layer architecture (perception, decision, action). Core hardware includes STM32 flight controller, OpenMV H7 Plus camera for visual perception, optical flow sensor for stable hovering, and custom 3D-printed structural components. The flight control system employs cascaded PID controllers for stable flight, which is fundamental to reliable autonomous control. The vision system uses classic computer vision algorithms such as color recognition in LAB color space and dynamic thresholding for tunnel navigation, obstacle avoidance, and target grasping. Most critically, modules are precisely coordinated through a custom state machine-based UART communication protocol.
The project's final result is a fully autonomous system that successfully completed 83.33% of overall tasks. Landing accuracy was better than 10 centimeters, and visual navigation detection accuracy exceeded 96%. The project also honestly documented challenges—tunnel traversal (20% success rate) and package pickup (60% success rate), reflecting the practical limitations of low-cost vision solutions in extremely complex environments. But it is precisely the success under these constraints that fully demonstrates that through innovative design and fine-tuning of core control algorithms, highly complex autonomous drone functions can be achieved under strict cost and hardware limitations.