Information-based geometric optimization model predictive control method and system for high dynamic trajectory tracking

CN123113393APending Publication Date: 2026-11-17CHONGQING UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611526976.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-29
Publication Date
2026-11-17

Smart Images

  • Figure CN123113393A_ABST
    Figure CN123113393A_ABST
Patent Text Reader

Abstract

This invention discloses an information geometry optimization model predictive control method and system for high-dynamic trajectory tracking. The method transcribes the finite-time optimal control problem into a probability measure space, parameterizing the sampling manifold with a family of diagonal Gaussian distributions. It derives the diagonal structure of the Fisher information matrix block of the diagonal Gaussian distribution, obtaining closed-form solutions of the natural gradients of the mean and variance, achieving adaptive joint updating of the mean and covariance of the sampling distribution on the statistical manifold. It replaces the exponential Softmax weights with Log-Rank weights based on relative sample ranking to eliminate the risk of weight collapse. It combines a time-domain rolling shift two-parameter hot-start strategy with an early stopping mechanism based on parameter stagnation rate, and implements a full-memory closed-loop parallel solution architecture using CUDA. This invention achieves geometrically consistent gradient-free optimization on the statistical manifold without requiring gradient information, significantly improving the accuracy, convergence quality, and real-time computational efficiency of quadrotor UAVs in high-dynamic racing trajectory tracking tasks.
Need to check novelty before this filing date? Find Prior Art