Framework and experiment for a constraints-defined longitudinal dynamics control for autonomous driving in traffic

This paper presents a practical constraint-defined control framework for longitudinal dynamics controllers for autonomous vehicles. The proposed framework aims at handling safety constraints, model error and uncertainty, as well as efficiency optimization. By including traffic context information, i.e., motion boundaries and harshness levels, the controller adapts its behavior to enhance safety and efficient vehicle operation, achieving a bounded tracking performance. Model predictive controller with numerical optimizer is chosen as the controller structure due to the multi-objective optimization nature and the requirement of handling real-time actuator and state constraints. Simulations and real-world driving tests are performed to demonstrate the effectiveness of the proposed control framework.