Model Predictive Control for Integrated Motion Planning for Control in Automated Vehicles

The aim of the project is to expand the basic methodology of Model Predictive Control with regard to autonomous driving. In numerous projects, MPC has already been identified as a promising approach for trajectory tracking control. The main advantage lies in its holistic and intuitive approach to controlling nonlinear multivariable systems with input and state constraints. However, for various reasons, MPC has not yet made its way into industrial systems. Primarily, MPC is known for its high computational effort and is also relatively complex to implement.

The research of this project aims at reducing these disadvantages and further strengthening the advantages. In particular, the approach of scenario-based MPC should be further expanded and reinforced. Additionally, the functional scope of MPC will be extended from pure trajectory tracking control to include preceding maneuver planning. This leads to a simplification of software architecture and reduction of existing interfaces as well as leveraging the strength of MPC in considering constraints and providing explicit safety guarantees for controlled system behavior. In order to avoid overly cautious vehicle behavior it's important to properly interpret & anticipate neighboring traffic behavior thus predicting future behaviors within SCMPC through scenarios. The SCMPC within the framework is used to predict the future behavior of other road users through scenarios. The basic idea is that the likely behavior as well as the inherent uncertainty are depicted by a set of scenarios. This approach is intuitive, computationally efficient, and easy to implement because empirical data can be processed directly. Furthermore, theoretical safety guarantees can be derived, which are to be expanded within the scope of the project.

In a second part of the project, a toolbox will be designed for enabling easy implementation of SCMPC on different systems. The toolbox should also help reduce computational complexity. To achieve this goal, efficient algorithms at state-of-the-art technology specifically tailored for combined maneuver planning and trajectory tracking control will be utilized. An integrated code generator aims to produce ready-to-use independent C code that can directly run on most electronic control units. A Graphical User Interface will facilitate operation and bypass many difficulties in implementation tasks. As a result, numerical issues will completely abstract from function development.

Overall, this project seeks to advance both current MPC technology with respect to automated driving and promote its industrial applicability.

The project is funded by the German Research Foundation (DFG), Project No. 460891204.