Walking deficits are a common bi-product of numerous neurological injuries, such as stroke and spinal cord injury. A number of new therapeutic interventions, such as body-weight supported locomotor training and robotic technologies aim to improve walking function and reduce co-morbidities. Currently, there is no way to determine what the optimal set of training parameters are for maximizing step performance. This paper presents a technique for estimating the walking performance of individuals with gait disorders using a robotic-orthosis. The device, called the Lokomat is coupled to the subject through instrumented leg cuffs, while the split-belt treadmill on which the subject walks is instrumented with piezo-electric force sensors allowing for the calculation of ground reaction forces and center of pressure. Using this data, a real-time inverse dynamics approach can be used to estimate the kinetics and kinematics of the subject, and when combined with electromyographic (EMG) data, the set of training conditions through which the subject generates the most appropriate EMG patterns and joint moments can be identified. The proposed technique will for the first time provide clinicians a way of determining the optimal gait training parameters for each individual, and also track their functional recovery throughout their neurorehabilitation program. It is postulated that training at the conditions that maximizes stepping performance will lead to higher gains in over-ground walking ability.