Dániel Pup

Gergő Ignéczi, Ernő Horváth, Dániel Pup

In this paper a self-developed controller algorithm is presented with the goal of handling a basic parking maneuver. One of the biggest challenges of autonomous vehicle control is the right calibration and finding the right vehicle models for the given conditions. As a result of many other research, model predictive control (MPC) structures have started to become the most promising control technique. During our work we implemented an MPC function from white paper. Considering the low-speed conditions of a parking maneuver we use a kinematic bicycle model as the basis of the controller. The algorithm has two main inputs: a planned trajectory and the vehicle state feedback signals. The controller is realized as a Simulink model, and it is integrated into a complete autonomous control system using ROS framework. The results are validated through multiple steps: using Simulink only with a pure kinematic bicycle plant model; using LGSVL simulation framework containing a real vehicle model and the entire software chain; the controller is prepared for real vehicle tests.

G. Rödönyi, G. I. Beintema, R. Tóth et al.

Automated driving applications require accurate vehicle specific models to precisely predict and control the motion dynamics. However, modern vehicles have a wide array of digital and mechatronic components that are difficult to model, manufactures do not disclose all details required for modelling and even existing models of subcomponents require coefficient estimation to match the specific characteristics of each vehicle and their change over time. Hence, it is attractive to use data-driven modelling to capture the relevant vehicle dynamics and synthesise model-based control solutions. In this paper, we address identification of the steering system of an autonomous car based on measured data. We show that the underlying dynamics are highly nonlinear and challenging to be captured, necessitating the use of data-driven methods that fuse the approximation capabilities of learning and the efficiency of dynamic system identification. We demonstrate that such a neural network based subspace-encoder method can successfully capture the underlying dynamics while other methods fall short to provide reliable results.