Near Real-Time Center of Gravity Estimation of an Unmanned Aerial Vehicle

As UAV technology advances and finds broader applications across industries, real-time estimation of the center of gravity (CG) becomes crucial for ensuring safe and efficient flight. Accurate CG estimation is vital not only for UAVs but also for manned aircraft, as it significantly influences stability and control by indicating the point where the entire weight is evenly distributed. Variations in CG, resulting from changes in altitude, attitude, speed, fuel consumption, and payload, directly impact stability and may lead to compromised performance and increased risk of accidents if not monitored. Existing methods focus on regression analysis, multi-accelerometer measurements, and pitching moment balance relationships; however, each is limited by dynamic changes and flight parameter uncertainties, affecting CG estimation accuracy and stability. This study presents a novel approach for near real-time CG estimation in UAVs by employing polar coordinate transformation and redefining transformations for acceleration components in the “instrument axes” system. By calculating and simulating the CG acceleration along each axis, the method provides an accurate representation of dynamic CG positioning. Weighted least squares were integrated to minimize noise and data inaccuracies, enhancing model robustness, while additional components like a forgetting factor, constraint equation, and gradient estimator refined and stabilized the estimation model. Results indicate that the weighted least squares method effectively manages data of varying magnitudes, maintaining errors within target ranges, essential for applications requiring consistent accuracy across operational scales. The model demonstrated reliable performance across multiple CG scenarios, with the y-axis achieving the lowest error rates and greatest stability, while minor adjustments to the x-axis could enhance accuracy for larger shifts. The z-axis showed a trend of reduced error with larger magnitudes, suggesting effective adaptation to more pronounced changes, though further fine-tuning could improve consistency across all magnitudes.