Electric vehicles with multiple motors permit continuous direct yaw moment control, also called torque-vectoring. This allows to significantly enhance the cornering response, e.g., by extending the linear region of the vehicle understeer characteristic and by increasing the maximum achievable lateral acceleration. These benefits are well documented for human-driven cars, yet limited information is available for autonomous/driverless cars. In particular, over the last few years, steering controllers for automated driving at the cornering limit have considerably advanced, but it is unclear how these controllers should be integrated alongside a torque-vectoring system. This contribution discusses the integration of torque-vectoring control and automated driving, including the design and implementation of the torque-vectoring controller of the autonomous and driverless electric racing vehicle for the RoboRace competition. The paper presents the main vehicle characteristics, together with the sensing system of the racetrack, and the control architecture. A quasi-static model is introduced that predicts the understeer characteristics at different longitudinal accelerations, and generates the moment method plots characterizing the performance envelope. The model is coupled with an off-line optimization for the a-priori investigation of the potential torque-vectoring benefits. The systematic computation of the achievable cornering limits is used to specify and design realistic maps of the reference yaw rate, and a non-linear feedforward yaw moment contribution to generate the reference cornering response in quasi-static conditions. A gain scheduled proportional-integral controller is adopted to increase yaw damping, thus enhancing the transient response. Results from simulation and experiments demonstrate the effectiveness of the proposed controller.