Separating rigid-body motion from elastic vibrations in launch vehicles is critical yet challenging, especially when the first-order elastic frequency is close to or coincident with the rigid-body control bandwidth. To address this, this paper proposes a novel control technology that leverages existing onboard sensors without requiring additional hardware. The key innovations are as follows: First, a signal differencing strategy using two inertial measurement devices at distinct locations is employed to eliminate rigid-body components physically, isolating pure elastic information. Second, a direct lattice adaptive notch filter (DLANF) with a forward–backward error-normalized stochastic gradient algorithm is applied to accurately track time-varying elastic frequencies. Third, based on these identified frequencies, a predefined-time nonsingular sliding mode observer is constructed to decouple the rigid-body and first-order elastic signals in the time domain. This approach overcomes the phase-lag limitations of traditional filters and the observability issues of conventional observers. Crucially, the method is validated not only through simulations but also using closed-loop data from an actual rocket flight, demonstrating its capability to identify weak low-order frequencies and recover clean rigid-body states for autonomous stability control.
Song et al. (Thu,) studied this question.