A Smart Landing Platform With Data-Driven Analytic Procedures for UAV Preflight Safety Diagnosis

Abstract

Due to the limitation imposed by hardware and form factor considerations, multirotor unmanned aircraft (drones) are unable to conduct preflight physical checks on their own capacity. Critical safety checks involve detecting various anomalies such as imbalanced payload, damaged propellers, mulfunctioning motors and poorly calibrated compass, etc. Human efforts are currently required for performing such tasks, which impedes large-scale deployments of drones and increases the operational costs. In this work, we propose a weight-measuring landing platform along with a set of statistical inference algorithms aimed at performing safety checks for any multicopter aircraft that lands on the platform. We develop a nonconvex nonlinear least squares model for estimating the center of gravity and orientation of the aircraft, and derive a recursive formula for calculating the optimal solution. In numeric experiments, our analytical solution method has been able to find the global solution orders-of-magnitude faster than a global optimization solver. We have conducted real-system tests on a quadcopter drone deliberately configured to carry misplaced payload, and to use damaged propellers. Experiment results show that the platform is able to detect and profile these common safety issues with high accuracy.

Keywords

• Anomaly detection
• nonlinear least squares
• statistical inference
• unmanned aircraft