Collision avoidance for autonomous vehicle using MPC and time varying Sigmoid safety constraints
Manel Ammour, Rodolfo Orjuela, Michel Basset
Abstract
This paper deals with the planning of collision-free trajectory for autonomous vehicle on highway. The problem is formulated using Model Predictive Control (MPC) where a simplified prediction model that integrates the relative positions and velocities between the surrounding vehicles and the ego vehicle is employed. The MPC provides the necessary lateral and longitudinal accelerations to the ego vehicle to track the reference path and to drive at the desired speed without hitting the road boundaries or colliding with the neighbouring vehicles. The collision avoidance is guaranteed through the integration of a time varying, Sigmoid-based, safety constraint. This constraint plays the role of an S-shaped barrier, enforcing the host vehicle to move apart and change the lane when the inter-vehicular safety distance is not kept. The proposed approach is focused on avoiding collision by performing lane change and overtaking maneuvers since in many traffic situations, an imminent collision cannot be avoided by braking. Satisfactory results are obtained from the set of the simulated scenarios.