Acceleration Controller Gain Adaptation for Vehicle Speed Tracking

The patent transforms the complex parameterization problem into a machine learning model training problem. Instead of manually parameterizing the acceleration controller for each vehicle model, the system collects operational data (speed setpoints, actual speeds, acceleration setpoints) and trains a machine learning model to automatically adapt the gain factor. This changes the approach from direct parameter specification to indirect parameter optimization through data-driven learning, thereby reducing parameterization complexity while maintaining or improving control precision.

The system enables self-adaptation of the acceleration controller by using its own operational data to train the machine learning model. The controller collects data during normal operation, the model is trained on this data, and the trained model automatically adjusts the gain factor without external intervention. This self-service mechanism eliminates the need for manual parameterization by engineers for each vehicle model, significantly reducing device complexity and parameterization effort.

The system performs preliminary data collection during normal operation, storing speed setpoints, actual speeds, and acceleration setpoints in buffers before the adaptation process is needed. This preliminary action allows the system to have ready-to-use training data when adaptation is required, reducing the processing time during critical adaptation phases while maintaining high adaptation accuracy through comprehensive data analysis.

The patent implements a two-stage adaptation approach: first, a coarse adaptation using a subset of data to quickly establish baseline parameters, then a finer adaptation using the complete dataset for precision optimization. This partial action strategy balances processing time and adaptation accuracy by not immediately processing all data at full detail,而是 progressively refining the gain factor with increasing data subsets.