Real-Time Control Allocation Method for Multi-Control-Surface Aircraft Based on Neighborhoodconstrained Local Linearization

Multi-control-surface aircraft, renowned for their stealth and maneuverability, face critical challenges in control allocation due to the absence of traditional stabilizers, which leads to control surface redundancy and nonlinear coupling (e.g., quadratic yaw moments in split drag rudders). Traditional linear pseudo-inverse methods, reliant on global control effectiveness matrices, suffer significant errors in nonlinear regimes, particularly in low-slope regions like drag rudder dead zones. To address this, we propose a real-time control allocation framework combining localized linear pseudo-inverse methods (LLPIM) with neighborhood constraints. The LLPIM replaces global matrices with localized Jacobian matrices at operating points and incorporates intercept terms to mitigate nonlinear effects. To prevent divergence in low-slope regions, deflection increments are constrained, ensuring validity of local linear approximations. Implemented within a six-degree-of-freedom multi-control-surface aircraft model with 10 control surfaces, simulations demonstrate that the constrained LLPIM reduces allocation errors by 3 times, achieving $<10\%$ error across 95% of the achievable moment set. Attitude tracking accuracy improves markedly, eliminating oscillatory instability in step responses and reducing residual errors for periodic commands. This low-complexity, robust framework offers a practical solution for real-time flight control of high-maneuverability multi-control-surface aircraft.