How is Model Predictive Control (MPC) used in autonomous vehicle control?

MPC optimizes control inputs over a prediction horizon while satisfying constraints. For autonomous vehicles: Vehicle model predicts future states given control inputs (steering, throttle/brake); cost function penalizes deviation from reference trajectory, control effort, and jerk; constraints include actuator limits, vehicle dynamics, comfort limits, and obstacle avoidance. Optimization solves for optimal control sequence, applying only the first input before re-optimizing. Advantages: Handles constraints explicitly, incorporates preview information (upcoming curves), optimizes multiple objectives. Challenges include computational load (real-time optimization), model accuracy, and tuning complexity. MPC is increasingly used for both longitudinal and lateral control in production systems.