Active Safety: Decision-Making using Potential Fields for Path-Following in Autonomous Driving

Motion planning is vital for autonomous driving as well as safely transporting passengers and goods. A core element of motion planning is decision-making and its implication for trajectory generation in the context of scenario-independent and time-dependent motion planning. This article presents a novel motion planning framework with an extended operational design domain (ODD) to shift boundaries towards scenario-independent decision-making for a given path, focusing on obstacle and crash avoidance. A nonlinear model predictive path-following controller (NMPFC) is utilized to obtain a decision with the related trajectory and track the trajectory simultaneously. This path-following framework is fed by an ego-motion prediction module and an environment representation via an artificial potential field (APF) used for optimal non-predefined decision-making. A weighted topographic environment model forces the NMPFC to avoid critical trajectories defined by their level of risk with respect to physical state and actuator constraints. The parameterization of this APF is challenging, such that a multi-layer definition and implementation of the APF for more accurate decision-making is required for the proposed active safety features. The optimizer’s cost function directly encapsulates and manages potential surrounding risks, while constraint relaxation maintains a limit on the computation time per optimization cycle. This synergistic interaction ultimately ensures recursive feasibility, which is mandatory for safe operation. Finally, we present a virtual testing environment consisting of Matlab/Simulink and IPG CarMaker to demonstrate the system’s abilities under different test cases.