Adaptive Control Allocation for Incremental Nonlinear Dynamic Inversion under Actuator Degradation

Incremental Nonlinear Dynamic Inversion (INDI) control is known to possess robustness properties with respect to modeling uncertainties, primarily due to its limited model dependence, but it can struggle to maintain performance under severe actuator degradation and, more generally, when the input map used for control allocation is highly inaccurate. Since control allocation is the only explicitly model-dependent element in INDI, this paper proposes and analyzes adaptive strategies to improve the accuracy of the allocation step. A predictorbased model reference adaptive control (PMRAC) approach is first introduced, followed by two online parameter-estimation methods: concurrent learning with directional forgetting and a two-layered recursive least-squares scheme. Simulation results on a nonlinear overactuated aircraft model demonstrate improved tracking performance compared with baseline INDI and the PMRAC augmentation under substantial input-map uncertainty, including severe actuator degradation.