Adaptive MPC Vehicle Route Planning

The patent divides the long time horizon into multiple shorter time horizons or segments. Instead of solving one large optimization problem over the entire route, the system solves a sequence of smaller MPC problems over shorter prediction horizons. This segmentation reduces the computational complexity of each individual optimization problem while collectively achieving coverage of the full time horizon, thus balancing optimality with computational feasibility.

The patent extracts and removes redundant or less significant inputs from the optimization problem. By identifying and eliminating unnecessary control variables or states that do not significantly impact the optimal solution, the system reduces the dimensionality of the optimization problem. This allows for higher sampling frequencies and better resolution without proportionally increasing computational burden, as the number of active optimization variables is reduced.

The patent implements dynamic adjustment of the time horizon length and sampling frequency based on operating conditions, vehicle state, and route characteristics. Rather than using fixed parameters, the system adapts the prediction horizon and sampling rate in real-time. This allows the system to use shorter horizons and lower sampling frequencies when computational resources are constrained or conditions are stable, while switching to longer horizons and higher frequencies when precision is critical, thus dynamically balancing computation speed with resolution quality.