This thesis studies integrated chassis control (ICC) for a four-wheel-steering and torque-vectoring (4WS+TV) electric vehicle, aiming to improve path tracking and stability while retaining real-time feasibility. Among ICC methods, nonlinear model predictive control (NMPC) is compelling for handling multivariable coupling, constraints, and look-ahead objectives, yet practical use depends on model fidelity, preview design, and computation. We implement two teacher controllers in CasADi–acados("NMPC-Simple" and "NMPC-DSTE")，each evaluated with/without preview (four controllers in total) and compared against classical lateral baselines (Stanley, PIDF2). The previewed DSTE configuration offers the best trade-off in cross-track/heading errors, yaw-rate behavior, and control smoothness; its preview gain is selected by sweep and it is adopted as the expert policy.