Fault-Tolerant Dynamic Control Allocation for Input Redundant Systems via Virtual Actuators

Dynamic allocators exploit actuator redundancy to distribute the control effort requested by a nominal controller among the available actuators. However, actuator faults, such as partial or complete loss of effectiveness, alter the plant input map and may invalidate the nominal allocation policy, leading to degraded closed-loop performance or even instability. To address this problem, we propose a plug-and-play fault-tolerant architecture for continuous-time systems with strong input redundancy. The nominal control-allocator loop is augmented with a Virtual Actuator (VA) that masks actuator faults and cancels the nominal allocator contribution. Concurrently, the VA reallocates the physical input within the post-fault null-space to reduce weighted input energy while guaranteeing state-invisibility. In particular, the post-fault reallocation gain is designed via a convex Linear Matrix Inequality (LMI) to minimize a certified upper bound on the weighted input energy. A numerical example validates the effectiveness of the approach.