© 2015 Elsevier Ltd. Increased practice difficulty during asymmetrical split-belt treadmill rehabilitation has been shown to improve gait outcomes during retention and transfer tests. However, research in this area has been limited by manual treadmill operation. In the case of variable practice, which requires stride-by-stride changes to treadmill belt velocities, the treadmill control must be automated. This paper presents a method for automation of asymmetrical split-belt treadmill walking, and evaluates how well this method performs with regards to timing of gait events. One participant walked asymmetrically for 100 strides, where the non-dominant limb was driven at their self-selected walking speed, while the other limb was driven randomly on a stride-by-stride basis. In the control loop, the key factors to insure that the treadmill belt had accelerated to its new velocity safely during the swing phase were the sampling rate of the A/D converter, processing time within the controller software, and acceleration of the treadmill belt. The combination of these three factors resulted in a total control loop time during each swing phase that satisfied these requirements with a factor of safety that was greater than 4. Further, a polynomial fit indicated that belt acceleration was the largest contributor to changes in this total time. This approach appears to be safe and reliable for stride-by-stride adjustment of treadmill belt speed, making it suitable for future asymmetrical split-belt walking studies. Further, it can be incorporated into virtual reality rehabilitation paradigms that utilize split-belt treadmill walking.