Shared Control Between Pilots and Autopilots: Illustration of a Cyber-Physical Human System
Although increased automation has made it easier to control aircraft, ensuring a safe interaction between the pilots and the autopilots is still a challenging problem, especially in the presence of severe anomalies. Current approach consists of autopilot solutions that disengage themselves when they become ineffective. This may cause reengagement of the pilot at the worst possible time, which can result in undesired consequences. In this paper, a series of research studies that propose pilot-autopilot interaction schemes based on the Capacity for Maneuver (CfM) concept, are covered. CfM refers to the remaining capacity of the actuators that can be used for bringing the aircraft to safety. It is demonstrated that CfM-based pilot-autopilot interaction schemes, or Shared Control Architectures (SCA), can be promising alternatives to the existing schemes. Two SCAs are tested in the experiments. In SCA1, the pilot takes over the control from the autopilot using the monitored CfM information. In SCA2, the pilot takes on the role of a supervisor helping the autopilot whenever a need arises, based on the CfM information. Whenever the aircraft experiences a severe anomaly, the pilot assesses the situation based on his/her CfM monitoring and intervenes by providing two control system parameter estimates to the autopilot. This increases the effectiveness of the autopilot. Using human-in-the-loop simulations, it is shown that CfM based interactions provides smaller tracking errors and larger overall CfM. The subjects including a commercial airline pilot and several university students are trained using a systematic procedure. Creation of new cyber physical & human systems is inevitable along with deeper engagement with the social science community so as to get better insight into human decision making. The results reported here should be viewed as a first of several steps in this research direction.
Code (0)
등록된 구현이 없습니다.
Tasks
Decision MakingSimilar Papers 제목 키워드 기반
Reinforcement Learning Based Prediction of PID Controller Gains for Quadrotor UAVs
A reinforcement learning (RL) based methodology is proposed and implemented for online fine-tuning of PID controller gains, thus, improving quadrotor effective and accurate trajectory tracking. The RL agent is first trai…
Reinforcement Learning (RL)A Data-Driven Autopilot for Fixed-Wing Aircraft Based on Model Predictive Control
Autopilots for fixed-wing aircraft are typically designed based on linearized aerodynamic models consisting of stability and control derivatives obtained from wind-tunnel testing. The resulting local controllers are then…
Model Predictive ControlSchedulingSystems Theoretic Process Analysis of a Run Time Assured Neural Network Control System
This research considers the problem of identifying safety constraints and developing Run Time Assurance (RTA) for Deep Reinforcement Learning (RL) Tactical Autopilots that use neural network control systems (NNCS). This …
Collision AvoidanceDeep Reinforcement LearningReinforcement Learning (RL)Cognitive Kernel: An Open-source Agent System towards Generalist Autopilots
We introduce Cognitive Kernel, an open-source agent system towards the goal of generalist autopilots. Unlike copilot systems, which primarily rely on users to provide essential state information (e.g., task descriptions)…
ManagementAcceleration-Based Control of Fixed-Wing UAVs for Guidance Applications
Acceleration-commanded guidance laws (e.g., proportional navigation) are attractive for high-level decision making, but their direct deployment on fixed-wing UAVs is challenging because accelerations are not directly act…
Decision Making