The team of engineers working on how the design and applications of electrical devices can be “more biological”, inspired by nature, developed a system that allows a robot to get out of a labyrinth with its own skills.
Most importantly, the software in the robot was able to remember previous wrong turns and correct its mistakes on the next ride. The robot managed to get out of the maze after 16 attempts to get out of a maze of about two square meters.
Inspired by brain structure
A team of engineers has developed a project that aims to let a robot learn the best route to get out of the maze on its own. The system, inspired by the neuromorphic circuit that records repetitive situations in the human brain and helps determine their preferences, was announced in an article published in the journal Science Developments. The researchers found a way to reduce the robot’s power needs and size by installing a neuromorphic circuit alongside software-based learning algorithms.
The circuit that made the robot work properly was composed of a substance called polymer, which mimics the brain. This item is able to preserve ‘experiences’ for a long time and use them to get out of the maze. The researchers stated that this substance can be defined as the memory of the labyrinth. Electrical engineer Imke Krauhausen from Eindhoven University of Technology said, “In an experiment, this is adjusted by applying a certain amount of electricity in our device, just as the direction-finding ability of a mouse trying to find its way in a maze is strengthened.” By adjusting the resistance in the device, the voltage controlling the motors is changed and thus it is determined whether the robot will turn right or left.
Systems like this one, which mimic the human brain, can go beyond helping robots to navigate and work independently, reducing dependency on software and the cloud. Stating that these smart devices can be integrated with real nerve cells due to their organic structure, Krauhausen said, “Let’s say you lost your arm in an accident. In such cases, you can use these devices to connect your body to a bionic hand.”
Video showing the robot’s learning process:
The designated path to the exit from the maze contained only clues that told the robot to turn left or right. The robot managed to get out of the maze on its own after 16 attempts.
32472
