In a previous paper we presented a neural network called FRONTAL that attempted to capture the ability of infants to actively self-organize in a non-stationary environment. This paper describes the procedure whereby FRONTAL is interfaced to a robotic system called DART. DART comprises of a PUMA-562 arm and a real-time image processing system. The implementation procedure consisted of first connecting FRONTAL to a simulated DART robot. After successful operation, FRONTAL was connected to the real hardware DART. The ability
of this robotic system to actively explore its surroundings and modify its behavior based on experience and environmental feedback confirms previous simulation results.