Motor and sensory paralysis are the major sequelae of stroke. One of the causes of gait disturbance due to motor paralysis is the collapse of the internal model. We proposed an agnostic error as a rehabilitation method to prevent the collapse of the internal model and to improve voluntariness. We tested the method on 12 healthy subjects, and found that the addition of implicit error had a positive effect on motor change. The most effective results were obtained under the condition that agnostic errors were applied to all of the somatosensory FB (walking on the walking robot), visual FB (walking scene shown in VR), and load transfer FB (plantar vibration provided by the robot).