Persona: Senior Robotics Engineer and Mechatronic Systems Analyst
Abstract:
This project documents the design, iteration, and final validation of a second-generation (V2) autonomous flying umbrella system. The primary engineering challenge transitioned from basic lift-generation (achieved in V1) to the integration of a robust autonomous tracking suite capable of following a human subject without manual pilot intervention.
The build utilized a heterogeneous compute stack comprising a professional-grade flight controller interfaced with a Raspberry Pi for real-time image processing and computer vision. Mechanical improvements focused on portability and structural rigidity, featuring a custom-designed central hub and folding arm mechanism manufactured via Fused Deposition Modeling (FDM) using Carbon Fiber (CF) Nylon. The development cycle faced significant technical hurdles, including sensor fusion failures with Indirect Time-of-Flight (iToF) cameras, hardware-level communication breakdowns, and the inadequacy of standard GPS for close-quarters precision. The final solution successfully integrated a high-speed camera tracking system and PID-tuned flight dynamics, demonstrating operational stability in simulated environmental conditions, including precipitation.
Technical Summary: Iterative Development of an Autonomous Aerial Umbrella System
0:00 Objective Re-specification: The project aims to upgrade the V1 manual drone-umbrella to a fully autonomous V2 system capable of following a target without a remote controller.
1:18 Prototyping Framework: To mitigate risk during the testing of tracking algorithms, a smaller standardized drone frame from HolyBro was used as a development platform for the flight controller and sensor suite.
2:12 Autonomous Compute Stack: The system architecture utilizes a Raspberry Pi for computer vision processing to identify the user's head, which then transmits spatial coordinates to the flight controller to maintain a centered position.
3:29 Mechanical Innovation: To address the cumbersome nature of V1, a folding arm mechanism with a locking plate was engineered. This allows the frame to transition from a compact "tripod" size to an operational quadcopter configuration.
5:00 Advanced Manufacturing: Structural components were 3D printed using CF-Nylon on a Bambu Lab H2D. The build leveraged a heated chamber to prevent warping and used dual-material printing for clean removal of complex support structures.
8:04 Flight Dynamics and Troubleshooting: Initial test flights revealed motor orientation errors and vibrations. Corrections required manual rewiring and software-level motor reversal to achieve lift.
9:39 Sensor Suite Iteration (iToF): The team initially attempted to use an indirect Time-of-Flight camera for depth sensing. This technology measures phase shifts in emitted light to calculate distance, though it encountered significant firmware errors (uncone errors) and connectivity issues.
11:18 Systemic Hardware Failure: A catastrophic failure occurred during high-altitude testing, resulting in a bricked Raspberry Pi and carrier board, necessitating a total hardware overhaul and pausing the project during the academic year.
14:09 Sensor Redundancy and Resolution: After failing to achieve precision with standard GPS (accurate only to 3 meters) and suffering further iToF camera failures, the project reverted to an upgraded high-speed camera tracking system.
15:45 Final Integration and Validation: After 358 days of development, the hardware was transferred to the full-scale umbrella frame. Final testing confirmed successful autonomous tracking and stabilization.
17:28 Environmental Resilience Test: The completed unit was subjected to heavy rain testing to validate its functional utility as a weather-protection device while maintaining flight stability.