It is well known that the performance of many antilock braking systems (ABS) deteriorate on rough, non-deformable surfaces due to a number of factors such as axle oscillations, wheel speed fluctuations and deficiencies in the algorithms. Rough terrain excitation further contribute to tyre problems such as loss of vertical contact and poor contact patch generation that leads to reduced longitudinal force generation. In this study, a slightly modified version of the Bosch ABS algorithm is implemented in Matlab/Simulink using co-simulation with a validated full vehicle ADAMS model that incorporate a valid high-fidelity FTire model. A non-ABS test vehicle is fitted with a commercial ABS modulator controlled by an embedded computer. The co-simulation model is validated with vehicle test data on both flat and rough terrains. Initial results show that wheel speed fluctuations on rough terrain cause inaccuracies in the estimation of vehicle velocity and excessive noise on the derived rotational acceleration values. This leads to inaccurate longitudinal slip calculation and poor control state decisions respectively. It is concluded that, although the correlation is not yet as desired, the combined use of a simulation model and test vehicle can be a useful tool in the research of ABS on rough terrains.