In this paper, vehicle cornering along roads with low friction coefficient is studied, and an autopilot design is proposed to satisfy desired handling performance. A novel hierarchical optimization approach is presented to generate offline solutions for vehicle cornering problem along roads with different friction coefficients and radii of curvature. A vehicle status definition is introduced as a function of vehicle states that contains data to indicate handling performance. At each control instant, vehicle statuses are calculated, and at the end of each scenario, they are clustered with k-means clustering technique. The vehicle status cluster centres are associated with the applied control commands. The pair formed by the vehicle status at cluster centre and the control command corresponds to a rule that produces the unique control command to be applied as a function of the vehicle status. The autopilot is constructed by a convex combination of these rules. This basic idea of autopilot design has been extended for motion along a specific rotation radius and friction coefficient; the control commands corresponding to arbitrary scenario parameters are obtained by a runtime scheduling of the weighted interpolation among the control commands corresponding to different scenario parameters. The proposed autopilot design is verified with vehicle motion under offline optimized control commands. The autopilot is also tested along roads with varying friction coefficient, and it is shown that the vehicle satisfies the desired handling performance.