Safety-constrained NMPC for autonomous vehicle lane keeping and lane change under disturbances

This paper presents a safety-constrained nonlinear model predictive control (NMPC) approach for autonomous vehicle (AV) lane keeping and lane change under disturbances. The controller is based on a nonlinear kinematic bicycle model and is implemented in a receding-horizon framework. To improve safety during lane-keeping and lane-change maneuvers, hard constraints are imposed on road boundaries, steering angle, vehicle speed, longitudinal acceleration, and lateral acceleration. In addition, conservative constraint tightening is introduced to improve tolerance to bounded uncertainty and small model errors. The proposed approach is evaluated in simulation in four phases: lane keeping, lane change, lane change with velocity disturbance, and lane change with combined velocity and steering disturbances. The simulation results show that the controller can track the desired lane, complete the lane-change maneuver, and maintain the imposed safety constraints in both nominal and disturbed cases for the scenarios considered. Across the four phases, peak lateral tracking errors remain below 0.05 m during nominal lane keeping, approximately 0.13 m during nominal lane change, and approximately 0.25 m under simultaneous dual disturbances, with road-boundary constraints satisfied at every simulation step in each phase. These results indicate that the proposed formulation provides a practical safety-oriented control framework for autonomous vehicle lane keeping and lane change in a simulation-based study using a kinematic bicycle model and margin-based constraint tightening.