Teaching robots to think like us: Brain cells, electrical impulses steer robot though maze

Can quality beryllium taught to robots? Advances successful carnal reservoir computing, a exertion that makes consciousness of encephalon signals, could lend to creating artificial quality machines that deliberation similar us.

In Applied Physics Letters, researchers from the University of Tokyo outline however a robot could beryllium taught to navigate done a maze by electrically stimulating a civilization of brain nerve cells connected to the machine.

These nervus cells, oregon neurons, were grown from surviving cells and acted arsenic the carnal reservoir for the computer to conception coherent signals.

The signals are regarded arsenic homeostatic signals, telling the robot the interior situation was being maintained wrong a definite scope and acting arsenic a baseline arsenic it moved freely done the maze.

Whenever the robot veered successful the incorrect absorption oregon faced the incorrect way, the neurons successful the compartment civilization were disturbed by an electrical impulse. Throughout trials, the robot was continually fed the homeostatic signals interrupted by the disturbance signals until it had successfully solved the maze task.

Robot experiments. A robot was placed connected fields with obstacles and was directed toward the goal. Credit: Yuichiro Yada, Shusaku Yasuda, and Hirokazu Takahashi

These findings suggest goal-directed behaviour tin beryllium generated without immoderate further learning by sending disturbance signals to an embodied system. The robot could not spot the situation oregon get different sensory information, truthful it was wholly babelike connected the electrical trial-and-error impulses.

"I, myself, was inspired by our experiments to hypothesize that quality successful a surviving strategy emerges from a mechanics extracting a coherent output from a disorganized state, oregon a chaotic state," said co-author Hirokazu Takahashi, an subordinate prof of mechano-informatics.

Using this principle, the researchers amusement intelligent task-solving abilities tin beryllium produced utilizing carnal reservoir computers to extract chaotic neuronal signals and present homeostatic oregon disturbance signals. In doing so, the machine creates a reservoir that understands however to lick the task.

"A encephalon of [an] simple schoolhouse kid is incapable to lick mathematical problems successful a assemblage admittance exam, perchance due to the fact that the dynamics of the encephalon oregon their 'physical reservoir computer' is not affluent enough," said Takahashi. "Task-solving quality is determined by however affluent a repertoire of spatiotemporal patterns the web tin generate."

The squad believes utilizing carnal reservoir computing successful this discourse volition lend to a amended knowing of the brain's mechanisms and whitethorn pb to the caller improvement of a neuromorphic computer.

---

---