Hybrid Robust Control and Reinforcement Learning for Optimal Upset Recovery

Dutoi, Brian; Richards, Nathan D.; Gandhi, Neha; Ward, Dave; Leonard, John B.

doi:10.2514/6.2008-6502

Cited by 28 publications

(5 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Research is also being done on various adaptive control methods for creating robust controllers meant to handle failures or upsets which push the aircraft beyond the normal flight envelope [35][36][37] .…”

Section: The Generic Transport Modelmentioning

confidence: 99%

Bifurcation Analysis of the Generic Transport Model with a view to Upset Recovery

Jaa

2012

AIAA Atmospheric Flight Mechanics Conference

View full text Add to dashboard Cite

There is a trend developing in research focusing on commercial aircraft safety, especially with respect to upset recovery and loss-of-control. In seeking to address the upset recovery problem from a control perspective and owing to the fact that upsets involve flight outside the regular flight envelope, the requirement for a wide-envelope non-linear model arises, along with the requirement for non-linear analysis techniques. These needs are addressed by NASA's Generic Transport Model (GTM) and bifurcation theory respectively, with the latter having been used extensively in the analysis of non-linear flight dynamics of military aircraft operating in broad flight envelopes. This paper focuses on the application of bifurcation theory to the GTM and includes an analysis of the initial results thereof. It covers the difficulties involved in combining the two tools, especially the changes that need to be made to the GTM simulation in order to make it compatible with the mathematics performed in bifurcation analysis. Bifurcation theory is applied primarily to the longitudinal dynamics of the aircraft to investigate the effects of the aircraft elevator and throttle on the aircraft behaviour. The results agree with other research that suggests the classical role these controls play in established upset recovery doctrine is effective, but often not correctly executed. The paper also investigates the coupling between the longitudinal and lateral dynamics of the aircraft, highlighting how aerodynamic stall negatively affects the lateral dynamics. NomenclatureGTM Generic Transport Model LOC Loss-of-Control UAV Unmanned Aerial Vehicle FTC Fault Tolerant Control V velocity magnitude α angle of attack β sideslip angle P x body-axis relative angular rate Q y body-axis relative angular rate R z body-axis relative angular rate Φ Euler 321 roll attitude angle Θ Euler 321 pitch attitude angle Ψ Euler 321 yaw attitude angle δ e absolute elevator deflection angle δ a absolute aileron deflection angle δ r absolute rudder deflection angle δ t absolute throttle percentage * PhD Student, Department of Electrical and Electronic Engineering, sjpauck@sun.ac.za. † Senior Lecturer, Department of Electrical and Electronic Engineering, jengelbr@sun.ac.za, AIAA Senior Member.

show abstract

Section: The Generic Transport Modelmentioning

confidence: 99%

Bifurcation Analysis of the Generic Transport Model with a view to Upset Recovery

Jaa

2012

AIAA Atmospheric Flight Mechanics Conference

View full text Add to dashboard Cite

show abstract

“…Research has mostly focused on designing reconfigurable controllers 12 , robust controllers 13 and adaptive controllers [14][15][16][17][18] that have the ability to accommodate or adapt to actuator failures and airframe damage. The families of adaptive controllers presented in Ref.…”

Section: Introductionmentioning

confidence: 99%

Bifurcation Analysis and Simulation of Stall and Spin Recovery for Large Transport Aircraft

Jaa

2012

AIAA Atmospheric Flight Mechanics Conference

View full text Add to dashboard Cite

The nonlinear dynamics of a large transport aircraft are analyzed to determine how the natural stability of the aircraft could be used for stall and spin recovery. The analysis is performed using the wide-envelope nonlinear simulation model for the Generic Transport Model (GTM). A bifurcation analysis is performed with respect to elevator to find the desirable stable equilibria of the aircraft inside the aerodynamic envelope, and to check for undesirable stable equilibria outside the envelope. For control surface deflections near zero, only one stable equilibrium branch is found, corresponding to the aircraft gliding with the steady-state angle of attack and sideslip angle well within the envelope. For excessive elevator deflections, two stable branches are found corresponding to stall and spin conditions. Simulations are performed to confirm the results of the bifurcation analysis and to analyze the transient behavior during the stall and spin departure and recovery. The results confirm that the aircraft enters stall and spin when excessive elevator is commanded, and recovers from stall and spin when the elevator is returned to zero deflection. The stall and spin recovery of a large transport aircraft can therefore be performed by setting the throttle to idle, returning the control surfaces to their neutral values, and relying on the aircraft's natural stability. The natural rate damping of the aircraft is sufficient to quickly recover the aircraft from spin, and the natural tendency of the aircraft to point its nose into the airstream is sufficient to recover the aircraft from stall. Nomenclature α = angle of attack, rad β = sideslip angle, rad P = roll rate, rad/s Q = pitch rate, rad/s R = yaw rate, rad/s AirSTAR = Airborne Subscale Transport Aircraft Research GTM = Generic Transport Model LOC = Loss-of-Control QLC = Quantitative Loss-of-Control Criteria UAV = Unmanned Aerial Vehicle

show abstract

“…Reinforcement learning strategies for optimal upset recovery have also been considered for uav applications. 16 In this approach large batches of simulation runs are used to train optimal recovery strategies for online use. As an alternative to the offline training based methods, constrained control approaches to the stall recovery guidance problem have been investigated.…”

Section: Introductionmentioning

confidence: 99%

Stall Recovery Guidance Using Fast Model Predictive Control

Schuet

Lombaerts

Kaneshige

et al. 2017

AIAA Guidance, Navigation, and Control Conference

View full text Add to dashboard Cite

Based on a detailed analysis of recent loss-of-control events, the Aircraft State Awareness Joint Safety Analysis Team has identified the need to develop algorithms and display strategies to provide control guidance for recovery from approach-to-stall or stall. In order to be effective, such guidance should enhance the pilot's ability to execute the Federal Aviation Administration's recommended stall recovery procedure. This paper explores the use of a fast model predictive control algorithm that determines near optimal recovery guidance, which quantifies the aircraft configuration and situation dependent recovery information required to maximize the effectiveness of the recovery. This information includes the magnitude of the initial pitch down maneuver, the specific amount of airspeed and thrust needed before pulling out of the recovery dive, as well as the maximum pitch-up rate that can be sustained without causing a secondary stall. The algorithm was integrated and tested with an in-house desktop simulator that implements the General Transport Aircraft model and the associated stall aircraft dynamics. Preliminary results are presented to demonstrate the use of the proposed approach as a recovery aid for pilots.

show abstract

Hybrid Robust Control and Reinforcement Learning for Optimal Upset Recovery

Cited by 28 publications

References 6 publications

Bifurcation Analysis of the Generic Transport Model with a view to Upset Recovery

Bifurcation Analysis of the Generic Transport Model with a view to Upset Recovery

Bifurcation Analysis and Simulation of Stall and Spin Recovery for Large Transport Aircraft

Stall Recovery Guidance Using Fast Model Predictive Control

Contact Info

Product

Resources

About