Physical Intelligence Based On Belousov-Zhabotinsky Reaction

Fish, birds, fireflies and other self-organization phenomena in mother nature promote the in-depth study of the self-organization theory. Oscillating chemical reaction in the self-organizing system is first discovered by B.P.Belousov and A.M.Zhabotinsky, named Belousov-Zhabotinsky (BZ) reaction after them, to demonstrate the real cyclical changes in the chemical wave pattern. Since this year with the study of BZ reaction in depth, gradually many practical problems in areas such as mathematics, chemistry, biology and graphics are related, and BZ reaction system has been extensive research and applications.Using the BZ reaction system of the two variables, through simulation we describe its dynamic behavior. Combined with the existing classical structures (T-Shaped coincidence detector, chemical diode), we designed a new unidirectional transmission structure and the cross-propagation structure, and verified them in simulation experiments. During the simulation process, we found that the process of chemical wave propagation in the BZ reaction is very similar to the pulse signal transduction in neurons, and also the process of pulse propagation in different configurations can be used to simulate the corresponding structures of neuron (cell body, sudden touch, etc.). These findings guide us to try to simulate neurons based on the BZ reaction by using reasonable design of geometry structure. Later, by encoding the propagating waves and structural adjustments we designed a full binary adder unit, and verified it through simulation. The unit can well be extended in plane, and coupled together to realize a two-bit, or even multi-bit binary adder. The implementation of chemical adder show that "complex" function can be achieved completely under the geometrically constrained structure based on the BZ reaction system, strengthening the confidence in building the neuron model. Then we gave a specific neuron model, and connect them with a reasonable way to form a neural network. The neurons system, in the form of chemical wave, reasonably explained the mechanism of firefly flash synchronization from the perspective of the neuron. In the study, the main contributions are presented as follows:(1) Through the simulation of the BZ reaction system, combined with the classic channel structures, we designed and verified a new unidirectional transmission structure and the cross-propagation structure; and based on the observation of simulation results of the BZ reaction with different configurations, we proposed a new idea that simulate the neurons through rational design of channel geometry based on the BZ reaction.(2) Based on the T-Shaped structure, unidirectional transmission structure and the cross-propagation structure, we first gave a complete chart of the chemical binary adder unit, and its effectiveness was verified through simulation, while noting that the chemical adder Is highly adaptable for actual experiment implementation.(3) Based on the BZ reaction system, a specific neuron model was performed by rational design of the geometry structure. Then the neurons were connected to form a neural network which can reasonably explained the mechanism of firefly flash synchronization. The system is the first theoretical human-engineered system that display physical intelligence completely under the geometry structure and the connections.