Safe Trajectory Optimization and Efficient-Offline Robust Model Predictive Control for Autonomous Vehicle Lane Change

Driving autonomous vehicles through diverse road
conditions at various high speeds poses a significant challenge.
To address this challenge, we propose a hierarchical control
strategy consisting of an optimization-based trajectory planner
in the first layer and an efficient-offline robust method employing
path tracking in the second layer. Considering vehicle parametric
uncertainties, the proposed hierarchical structure addresses
multiple scenarios with varying road surface conditions and
velocities. In the first layer, using the Pontryagin maximum
principle (PMP) flexibly with the time-to-collision (TTC) method,
the motion planner generates a safe-optimal lane-change trajectory
when interacting with forward vehicles and adjacentlane
vehicles to improve ride comfort while maintaining safe
distances. In the second layer, the efficient offline robust model
predictive control (RMPC) with terminal constraints is applied
to a linear parameter varying (LPV) system. Utilizing linear
matrix inequality (LMI) techniques, the optimization problem
accommodates parametric uncertainties while robustly satisfying
input and output constraints. To emphasize superior performance,
we have considered comparing our proposed approach
with several state-of-the-art methods. Therefore, comparative
simulation results have shown that our safe-optimal trajectory
generation is better than the Spatio-Temporal Corridors method
regarding path smoothy. The proposed approach (i.e., efficientoffline
RMPC) then outperforms the offline MPC method in
terms of path-tracking performance and parametric uncertainty
handling while outperforming the online RMPC method in terms
of computational complexity reduction. Further, all methods are
verified using a co-simulation and testing platform via a highfidelity
dynamics testing vehicle control software (i.e., CarSim).