Autonomous Vehicle Trajectory Optimization Fallback Mechanism

The patent implements a fallback mechanism that prepares an alternative finite element based optimization method in advance. When the spline curve based QP optimization fails to converge, the system automatically switches to the finite element method, which uses a different mathematical formulation that is more numerically stable. This beforehand preparation of an alternative solution cushiones against the reliability issues of the primary optimization method.

The patent changes the mathematical parameters and formulation of the optimization problem by offering two different approaches: spline curve based QP optimization and finite element based optimization. When numerical issues arise in one parameterization, the system switches to different parameters in the alternative formulation, maintaining optimization precision while avoiding convergence failures.

The patent implements a dynamic optimization system that can switch between different optimization methods based on real-time convergence status. The system starts with the spline curve based method and dynamically transitions to the finite element method only when needed. This dynamic approach improves reliability while minimizing the actual complexity introduced, as the alternative method remains dormant until required.

The patent introduces a convergence check mechanism as an intermediary that monitors the optimization process and decides when to switch methods. This intermediary layer manages the complexity by providing a clear decision boundary: if convergence is achieved, use the primary method; if not, switch to the fallback method. This structured approach to complexity management makes the system more tractable.

The patent performs preliminary action by establishing the finite element based optimization method and convergence checking mechanisms in advance. However, the actual computational work is only performed by the primary spline curve based method unless convergence fails. This preliminary preparation ensures robustness while minimizing computational time, as the fallback method is ready but not executed unless absolutely necessary.

The patent implements a skipping strategy where the system rushes through the primary optimization method first, checking for convergence at each iteration. Only when convergence failure is detected does the system skip to the alternative finite element method. This approach minimizes computational time by avoiding unnecessary execution of the fallback method while maintaining robustness through the ability to switch.