In this paper, the reflexive behavior of a biomorphic adaptive robot is analyzed. The motion generation of the robot is governed by a Reaction-Diffusion Cellular Neural Network (RD-CNN) that evolves towards a Turing pattern representing the action pattern of the robot. The initial conditions of this RD-CNN are given by the sensor input. The proposed approach is particularly valuable when the number of sensors is high, being able to perform data compression in real-time through analog parallel processing. An experiment using a small 6-legged robot realized in Lego MindStorms™ with three sensors is presented to validate the approach. A simulated 3×3 CNN is used to control this hexapod.