Automatic Loss of Control Prevention using Longitudinal and Lateral Control for Run-time Assurance in General Aviation

Improving safety for aviation has been a fundamental and global objective for decades. General aviation aircraft collision with terrain causes a majority of serious injuries and fatalities. Uncontrollable terrain collision caused by a poor energy state and loss of aerodynamic control leads to Loss of Control Inflight (LOC-I) which accounts for 40% of total general aviation fatal incidents. The primary component of loss of control prevention (LOCP) systems is the ability to postpone the onset of LOC through flight envelope protection and limitations. This protection extends to inadvertent events with high angle of attack exceedance, abnormal aircraft attitude, high angular rates and accelerations, airspeeds, and flight trajectories. The emphasis for LOCP is on detect and avoid (prevent) rather than LOC recovery maneuvers. A specific LOCP system must consider an aircraft’s performance constraints and nominal flight envelope to determine appropriate LOC prevention controls for that aircraft. These LOCP system considerations become increasingly more important when the LOCP system is automatic (Auto LOCP), multiple safety systems may be present, and safety must be assured, especially for general aviation with a pilot in command. If general aviation terrain collision is to be successfully mitigated by an Auto LOCP system, the envelope protection and control systems must be systematically tested, evaluated, and verified across a wide range of possible flight envelope conditions to provide necessary Auto LOCP benchmarks for autonomy and safety. This paper proposes a framework utilizing a multi-modal Auto LOCP system using independent longitudinal and lateral flight controls, integrated with a run-time assurance (RTA) architecture, to successfully prevent LOC in general aviation aircraft. This research is conducted in a real-time six degree of freedom (6DoF) flight simulation using a Cessna 172 aircraft model.