Constrained chaos in three-module neural network enables to execute multiple tasks simultaneously

Constrained chaos introduced into a three-module neural network having feedforward inter-module structure could have potential abilities to execute multiple tasks simultaneously. Each module consists of a large number of binary state (±1) neurons. The entire activity pattern (neuron state) is updated by recurrent rule under certain external input to the first module and input to post-module from pre-module. As a practical example, with use of computer experiments, the proposed idea is applied to a robot actuator in which control system using chaos is installed. The three modules are assigned to the sensory neuron module, the inter neuron module, and the driving (motor) neuron module, respectively. Initially, the actuator system of robot is designed so as to generate the four different kinds of specific driving signals in the motor module via the interneuron module corresponding to the four specific inputs to the entire sensory neurons. Next, chaos is introduced by reducing connectivity in intra-modules and/or inter-modules as well. It results in generating of chaotic motion signals from the motor module. Third, when two fragment inputs which belong to any two of the four specific inputs are applied simultaneously, then the motor module gives corresponding two driving signals simultaneously. Nevertheless, chaotic activities are kept even if strong two fragment inputs to the sensory module are applied. The results are one of the typical examples to show that constrained chaos in neural systems having big redundancy is able to execute multiple tasks simultaneously as brain does.