Autonomous vehicle trajectory following with flatness-based model predictive control and real-time iteration

Model Predictive Control (MPC) is increasingly gaining popularity in a wide range of autonomous vehicle and driving assistant applications owing to its style of optimality and convenience in handling multiple constraints. However, a primary drawback in online applications lies in the high computational burden caused by the complex non-linearity in the vehicle system and driving stability constraints. The differential flatness property of the vehicle motion system offers a solution by treating the error signals as flat outputs for endogenous feedback linearization. Nevertheless, this approach introduces a new challenge of dealing with time-varying implicit constraints. To address this issue, our work combines flatness-based MPC design with real-time iterative handling of lateral stability constraints, leading to a pure linear Quadratic Programming (QP) problem that can be efficiently solved online. Through numerical simulations conducted under two challenging scenarios, we demonstrate that our proposed method outperforms conventional linear-time-varying MPC in terms of linearization error and dynamic performance. Finally, field tests conducted with a four-wheel independent driving and four-wheel independent steering chassis platform validate the feasibility and substantial potential of our presented method in autonomous driving applications.