Implementation and Validation of a Thrust-Mass-Acceleration Consistency Invariant for Quadcopter UAV Safety

Abstract

This paper introduces a thrust-mass-acceleration consistency invariant, a physics-based supervisory validation technique that utilizes Newton's Second Law to detect discrepancies in vertical thrust, mass, and acceleration during autonomous aerial vehicle operation. As most flight stacks currently utilize sensors and software to manage their operations, they generally assume that all sensor data received by the vehicle is accurate and trustworthy. This invariant was developed to provide a lightweight supervisory monitoring system that does not require modification of the flight stack but can detect when there is a discrepancy in one or more of the physical variables required for safe vertical flight. It uses the PX4 SITL with ROS2 integration to implement its supervisory validation capability within the constraints of a strictly defined North-East-Down frame of reference. In contrast to many existing approaches, this invariant leaves the control logic unchanged. Therefore, this invariant provides a modular framework for developing supervisory safety architectures based upon invariant relationships. Simulation studies have demonstrated that this invariant can effectively detect faults in the vertical thrust generation system and inconsistent vertical motion, while restricting its scope strictly to vertical dynamic consistency.