A novel road adaptive Linear Parameter-Varying (LPV) based controller for the semi-active suspension system of an automotive vehicle is proposed. The analysis is carried on the front-left Quarter of Vehicle (QoV) model generated via CarSimTM vehicle simulator. By using an on-line road roughness estimation, considered as a scheduling parameter, the proposed LPV/H? controller is designed to improve comfort and road holding. The road profile detector is based on the frequency and amplitude estimation of the road irregularities by using a Fourier analysis. An H? robust observer is designed to estimate the variables related to the QoV vertical dynamics, which are used to compute the road frequency and roughness. Different ISO road classes are used to evaluate on-line the proposed road identification algorithm. A Receiver Operating Characteristic (ROC) curve is used to monitor the performance of the roughness estimation; the results show that any road can be identified (at least 70% of success with a false alarm rate lower than 5%). The average error of road identification is 16.2%. Simulation results show that the proposed controller with road adaptation is capable to manage the trade-off between comfort and road holding. The road adaptive controller increases the comfort (35.8%) when the vehicle is driven on a road of bad quality, by considering an uncontrolled damper (passive suspension) as a benchmark. If the vehicle is driven on a smooth runway at high velocity, the proposed controller improves the road holding around 50%.