Kalman Filter Based Adaptive Control Allocation

In recent years, various approaches for adaptive control allocation have been published with the goal to enable fault tolerance and improve flight performance of overactuated vehicles. Most of the presented methods are based on Model Reference Adaptive Control (MRAC) techniques and/or recursive least squares. In this study, we present a novel strategy combining concurrent learning MRAC and Kalman filtering to adapt the control effectiveness vector. The proposed strategy incorporates short term updates of the direct update as well as stationary correction through the Kalman filter. The adaptive control allocation has been evaluated using a hexacopter high fidelity simulation model. The hexacopter is controlled by an Incremental Nonlinear Dynamic Inversion (INDI) velocity controller and by a backstepping controller. In the simulation, we demonstrated the advantages of the adaptive control allocation for partial loss of effectiveness of one rotor and in nominal conditions. The adaptive control allocation avoided the crash of the hexacopter and improved tracking in forward flight compensating for effectiveness changes.