Automobile drivers/passengers perceive automatic transmission (AT) shift quality through the torque transferred by transmission output shaft, so that torque regulation is critical in transmission shift control and etc. However, since a physical torque sensor is expensive, current shift control in AT is usually achieved by tracking a turbine speed profile due to the lack of the transmission output torque information. A direct torque feedback has long been desired for transmission shift control enhancement. This paper addresses a “virtual” torque sensor (VTS) algorithm that can provide an accurate estimate on the torque variation in the vehicle transmission output shaft using (existing) speed sensors. We have developed the algorithm using both the transmission output speed sensor and anti-lock braking system speed sensors. Practical solutions are provided to enhance the accuracy of the algorithm. The algorithm has been successfully implemented on both FWD and RWD vehicles. Compared to the measurement from a physical torque sensor, the torque estimation from the VTS algorithm is consistent and robust under real-world driving conditions. This technology can be applicable to front wheel drive (FWD), rear wheel drive (RWD) and all wheel drive (AWD) transmissions, and provide a low-cost solution to torque-based powertrain control enhancements.