The memristive brain project is dedicated to the reimplementation of bio-plausible emotional drives based behavioral strategies in a robotic system, in the other words to solve the embodiment problem for the simulated brain in HPC system.
During the NeuCogAr project we have managed to develop mammalian emotional drives potential basement for decision making over behavioral strategies. We are doing the validation of overall model of 8 basic emotional states: fear, joy, disgust, humiliation, anger, interest, surprise, distress based on three neuromodulators: noradrenaline, dopamine and serotonin and currently we have managed to validate two states: fear-like and disgust-like.
The problem is that the bio-plausible simulation of a mammalian brain is really slow and can not be used for real-time processing, but robotic embodiment should be operating real-time. We have started the Robot Dream project to solve this problem still keeping bio-plausibility in integration with real-time robotic system.
A robotic system management system could be implemented using several approaches including reimplementation of mammalian brain structures in electronic schematic. To implement this we have started from the memristive solution of a neuron. Starting from the beginning of 2017 we are developing a memristive implementation of bio-plausible neuronal inhibition and neuromodulation.
The complete documentation on memristive approach could be found here.
AI and robotics: New architectures of robotic systems, based on the integration of memory and processor in a single chip, capable of memorizing the information and changing accordingly connections between elements within processor will be implemented. This system will be significantly different from the existing computers/robots. The proposed architecture of new generation of robots will be capable of self learning and will not be using traditional software approaches making software development obsolete.
Brain science/neurophysiology: In the case of success, neurophysiology will acquire a new tool for modeling of processes in nervous system and brain. An integration of living (neurons) and not living memristive elements will be combined in a hybrid system.This tool will provide an option to better understand processes of the reinforcement and inhibition of signal pathways during supervised and unsupervised learning. Memristors eliminate all physical restrictions of learning algorithms; they allow to “switch off” learning of selected elements, allowing the creation of new neuromorphic machine learning algorithms.
From the bionics perspective, memristor-based artificial synapses could be a part of implantable controllers for artificial organs and prosthesis. Apart from the breakthrough in the main direction of the research, mentioned above, we expect also to have fundamental results in the following branches:
Physics/electronics: The project results will allow to better understand processes, occurring in different memristive devices – new perspective electronic elements. Technological processes of the production of such elements and circuits, as well as processors – memory elements based on them will be developed. We suppose that the investigation and development of memristive devices could result to the revolution in electronic and computer industries. “Being combined with transistors into hybrid chips, memristors could improve radically properties of digital circuits, without the removal of transistors. More effective utilization of transistors could allow to improve properties according to the Moor low, at least during next 10 years, without expensive and becoming continuously more complicated increasing of the integration level in the chip. Finally, memristors could be even a basic stone of new analog circuits, making calculations using the architecture, similar to that of the brain”. [William R.S., How We Found the Missing Memristor. IEEE Spectrum 2008-12-18]. Taking into consideration a grate and continuously increasing interest of scientific centers and commercial laboratories to the implementation of memristors, it is possible to predict that in 5-10 years memristors will be one of the most common elements of electronic circuits.
Material science/self organization: it will be investigated new polymer-based materials, allowing to organize structures with distributed electrical properties for the utilization as matrices with adaptive properties.
- demo-stand - contains info about current measuremet schematic. Now not used.
- doc - contains all required documentation
- ltspice_libs - contains all additional components, used in LTSPICE schematics
- memristors - contains information about memristors, used in project
- test-circuits - contains all electrical schematics, connected with this project
This video contains all info, related to current circuits.