Mechanical impedance compensation is very important for precise human-robot interaction and safety of wearable robots. The high-level assistive torque control must precede the zero-impedance system guarantee; therefore, the zero-impedance control loop must be faster than the high-level torque control loop. However, the process requires a large computing power and causes a CPU load on the central controller. As a result, it limits the real-time operation of high-level algorithm. In this paper, a system architecture that compensates mechanical impedance through motor driver network instead of a central controller is introduced. By applying this architecture to an actual robot platform and conducting an experiment, it was confirmed that the mechanical impedance decreased and the CPU load of the central controller was reduced.