Adaptive Model Predictive Control for Autonomous Vehicle Path Tracking

The patent implements dynamic operational parameters that are continuously updated based on real-time vehicle conditions and external factors. The machine learning circuit dynamically adjusts parameters such as vehicle mass, friction coefficients, and aerodynamic drag based on sensor inputs and simulated experiences, transforming the static controller into an adaptive system that evolves with changing conditions.

The system employs self-service through automated machine learning processes that continuously train and update the operational parameters without manual intervention. The neural network automatically processes sensor data, simulates vehicle behavior, and refines parameters based on performance feedback, enabling the controller to self-optimize its performance across diverse operating conditions.

The system performs preliminary action by pre-training the machine learning model through extensive simulations before actual vehicle operation. The neural network is trained offline using simulated vehicle behavior across various conditions, allowing it to learn optimal parameter values in advance. This pre-learning reduces the computational burden during real-time operation while maintaining high path tracking accuracy.

The patent uses copying by creating virtual copies of the vehicle system in simulation environments. The machine learning model learns from simulated vehicle behavior and external conditions, copying real-world physics and dynamics into the virtual training environment. This allows the system to accumulate training data and refine parameters without requiring extensive real-world testing, reducing overall system complexity.