Adaptive MPC Parameter Updates for Autonomous Vehicle Path Tracking

The patent applies dynamics by transitioning from fixed parameters to dynamically adjustable parameters. The machine learning model learns optimal parameter values during simulation and updates them in real-time based on changing vehicle conditions (tire wear, mass changes) and external conditions (road friction, weather), enabling the controller to adapt its behavior dynamically rather than relying on static preset values

The patent implements parameter changes by using a machine learning model to continuously adjust operational parameters such as prediction horizon, control horizon, sampling time, and MPC weights based on learned relationships between vehicle state, environmental conditions, and optimal control performance. This allows the controller to modify its parameters adaptively without changing its fundamental structure

The patent applies preliminary action by pre-training the machine learning model offline using extensive simulations that cover a wide range of vehicle and environmental conditions. This pre-training phase allows the system to learn optimal parameter relationships beforehand, so that during real-time operation, the model only needs to infer parameters based on current sensor inputs without requiring complex online computation or retraining

The patent uses copying by creating a virtual training environment that replicates real-world vehicle dynamics and conditions. The machine learning model is trained in this simulated copy of the real system, allowing extensive experimentation and learning without risking actual vehicle safety or performance. The learned parameters are then transferred to the real vehicle control system