Piecewise trajectory replanner for highway collision avoidance systems with safe-distance based threat assessment strategy and nonlinear model predictive control

Abstract

This paper proposes an emergency Trajectory Replanner (TR) for collision avoidance (CA) which works based on a Safe-Distance Based Threat Assessment Strategy (SDTA). The contribution of this work is the design of a piecewise-kinematic based TR, where it replans the path by avoiding the invisible rectangular region created by SDTA. The TR performance is measured by assessing its ability to yield a maneuverable path for lane change and lane keeping navigations of the host vehicle. The reliability of the TR is evaluated in multi-scenario computational simulations. In addition, the TR is expected to provide a reliable replanned path during the increased nonlinearity of high-speed collisions. For this reason, Nonlinear Model Predictive Control (NMPC) is adopted into the design to track the replanned trajectory via an active front steering and braking actuations. For path tracking strategy, comparisons with benchmark controllers are done to analyze NMPC’s reliability as multi-actuators nonlinear controller of the architecture to the CA performance in high-speed scenario. To reduce the complexity of the NMPC formulation, Move Blocking strategy is incorporated into the control design. Results show that the CA system performed well in emergency situations, where the vehicle successfully replanned the obstacle avoidance trajectory, produced dependable lane change and lane keeping navigations, and at the same time no side-collision with the obstacle’s edges occurred. Moreover, the multi-actuators and nonlinear features of NMPC as the PT strategy gave a better tracking performance in high-speed CA scenario. Assimilation of Move Blocking strategy into NMPC formulation lessened the computational burden of NMPC. The system is proven to provide reliable replanned trajectories and preventing multi-scenario collision risks while maintaining the safe distance and time constraints.