A noise-induced stability in the real-time robotic system for object handling

In various engineering fields, separation between signal and noise is one of the important issues for the robustness of systems that are working in the real environment, and the noise reduction has been discussed when designing a robust system. On the other hand, some researches reported that adding a noise contributes to having multiple internal states and enhancing a transition between the states in the case of non-linear and biological systems, such as the stochastic resonance. It leads to a hypothesis of neuro-mimetic models in which the noise enhances their performance. We have developed the robotic platform as a combination between the real-time simulator of neural dynamics and the robotic device operating in the real world. According to communicative interruptions and time lags, the real-time simulator has the limitation in ability to control the robot, especially in time domain. The robot frequently fails in making an action with respect to the previous sensor data if the calculation is done in the proper timing, providing a deadlock behavior. We here investigated the effect of the noise induction for escaping the deadlock and completion of the ball-handling task, and reported that a self-biased noise helps a enlargement of the range of delay in the system for exhibiting proper performances. By focusing on the temporal aspect of the noise effect in non-linear systems, our research approach may benefit to the implementation and development of biological models in the real system.