It is often quoted that the processor is the brain of the computer. But a processor works fundamentally differently from a human brain. Transistors use electronic signals to perform logical operations. Instead, the brain works with nerve cells, called neurons, which are connected by biological transmission pathways called synapses. At a higher level, the brain uses this signal to control the body and perceive its surroundings. When certain stimuli are perceived, responses from body/brain systems — for example, through the eyes, ears, or touch — are triggered through the learning process. For example, children learn not to reach for a hot stove twice: an input stimulus leads to a learning process and a definite behavioral outcome.
Working with the scientists, Paschalis Gkoupidenis, team leader in Paul Blom's department at the Max Planck Institute for Polymer Research, This basic principle of learning through experience has now been reduced to a form, using so-called organic neuromorphic circuits to guide robots through mazes. The work is an extensive collaboration between Eindhoven University, Stanford University, The University of Brescia, the University of Oxford, and KAUST.
"We wanted to use this simple device to demonstrate the power of this' organic neuromorphic device 'in the real world," Imke Krauhausen, a doctoral student in the Gkoupidenis group of the Van de Burgt Group and lead author of the scientific paper.
To navigate the robot through the maze, the researchers fed sensory signals from the environment into an intelligent adaptive circuit. The maze's path to the exit is visually displayed at the intersection of each maze. In the beginning, robots tend to misinterpret visual signals, making the wrong "turn" decision at the intersection of the maze and losing their way out. When the robot makes these decisions and goes down the wrong cul-de-sac, it is dissuaded from making the wrong decision by a corrective stimulus. Corrective stimuli, such as when the robot hits a wall, are applied directly to an automatic circuit via an electrical signal sensed by a touch sensor attached to the robot. With each subsequent experiment, the robot gradually learned to make the right "turn" decision at an intersection — avoiding corrective stimuli — and found its way out of the maze after a few trials. This learning process occurs only on organic adaptive circuits.
"We were very excited to see that the robot was able to navigate a maze after just a few laps by learning a simple organic circuit. Here we show a very simple initial setup. In the distant future, however, we expect organic neuromorphic devices to also be used for local and distributed computing/learning. This will open up entirely new possibilities for real-world applications in robotics, human-machine interfaces, and real-time diagnostics. At the intersection of materials science and robotics, new platforms for rapid prototyping and education are also expected to emerge."Gkoupidenis said.
