The development of children’s cognitive and perceptual skills depends heavily on object exploration and manipulative experiences. New types of robotic assistive technologies that enable children with disabilities to interact with their environment, which prove to be beneficial for their cognitive and perceptual skills development, have emerged in recent years. In this study, a human-robot interface that uses Event-Related Desynchronization (ERD) brain response during movement was developed. A haptic robot generates force feedback to constrain the user’s hand motion into a defined region using a Forbidden Region Virtual Fixture (FRVF). Two channels of electroencephalography (EEG) brain signals were acquired and used to design classifiers to discriminate the pattern of brain signals between “movement” and “rest” during the operation of the robot. The highest ERD classification accuracy achieved was 69.4%. With the classifier, ERD-based FRVFs were tested in a simple robot operation task with two non-disabled adults. The virtual wall triggered by ERD successfully blocked the motion of the two participants on average 7 and 6.4 out of 10 times, respectively. Although further improvement of the system is still needed, the experimental results demonstrate the potential of the ERD-based FRVFs in robot control for a clinical population.