During space missions, real time tele-operation of a rover is not practical because of significant signal latencies associated with inter planetary distances, making some degree of autonomy in rover control desirable. One of the challenges to achieving autonomy is the determination of terrain traversability. As part of this field, the determination of motion state of a rover on rough terrain via the estimation of wheel-terrain contact angles is proposed. This thesis investigates the feasibility of estimating the contact angles from the kinematics of the rover system and measurements from the onboard inertial measurement unit (IMU), joint angle sensors and wheel encoders. This approach does not rely on any knowledge of the terrain geometry or terrain mechanical properties. An existing framework of rover velocity and wheel slip estimation for flat terrain has been extended to additionally estimate the wheel-terrain contact angle along with a side slip angle for each individual wheel, for rough terrain drive. A random walk and a damped model are used to describe the evolution of the contact angle and side slip angle over an unknown terrain. A standard strapdown algorithm for the estimation of attitude and velocity from IMU measurements, is modified to incorporate the 3D kinematics of the rover in the implementation of a nonlinear Kalman filter to estimate the motion states. The estimation results from the filter are verified using tests performed on the ExoMars BB2. The obtained contact angle estimates are found to be consistent with the reference values.