Background: Collaborative robots are used in rehabilitation and are designed to interact with the client so as to provide the ability to assist walking therapeutically. One such device is the KineAssist which was designed to interact, either in a self-driven mode (SDM) or in an assist mode (AM), with neurologically-impaired individuals while they are walking on a treadmill surface. To understand the level of transparency (i.e., interference with movement due to the mechanical interface) between human and robot, and to estimate and account for changes in the kinetics and kinematics of the gait pattern, we tested the KineAssist under conditions of self-drive and horizontal push assistance. The aims of this study were to compare the joint kinematics, forces and moments during walking at a fixed constant treadmill belt speed and constrained walking cadence, with and without the robotic device (OUT) and to compare the biomechanics of assistive and self-drive modes in the device. Method: Twenty non-neurologically impaired adults participated in this study. We evaluated biomechanical parameters of walking at a fixed constant treadmill belt speed (1.0 m/s), with and without the robotic device in assistive mode. We also tested the self-drive condition, which enables the user to drive the speed and direction of a treadmill belt. Hip, knee and ankle angular displacements, ground reaction forces, hip, knee and ankle moments, and center of mass displacement were compared "in" vs "out" of the device. A repeated measures ANOVA test was applied with the three level factor of condition (OUT, AM, and SDM), and each participant was used as its own comparison. Results: When comparing "in" and "out" of the device, we did not observe any interruptions and/or reversals of direction of the basic gait pattern trajectory, but there was increased ankle and hip angular excursions, vertical ground reaction force and hip moments and reduced center of mass displacement during the "in device" condition. Comparing assistive vs self-drive mode in device, participants had greater flexed posture and accentuated hip moments and propulsive force, but reduced braking force. Conclusions: Although the magnitudes and/or range of certain gait pattern components were altered by the device, we did not observe any interruption from the mechanical interface upon the advancement of the trajectories nor reversals in direction of movement which suggests that the KineAssist permits relative transparency (i.e. lack of interference of movement by the device mechanism) to the individual's gait pattern. However, there are interactive forces to take into account, which appear to be overcome by kinematic and kinetic adjustments.