Skip to content

Gottingen robot learns how best to move on its own

Following up on the report from researchers at the Oregon State University who hope to use cockroaches as inspiration for developing more mobile robots, scientists at the Bernstein Center for Computational Neuroscience Göttingen have completed work on a robot that can change different gaits on its own. This is, again, directly taken from the animal world, where insects use different gaits when crawling up or downhill.

Central Pattern Generators (CPGs) that automatically control breathing and movement are found in many animals, including humans. Robots in the past were often developed using one CPG for every gait, which means they collect info about their surroundings from their sensors and then decide how best to proceed over the obstacle or terrain ahead.

The Gottingen scientists, however, have much improved the process by designing their robot to use a single CPG that automatically chooses and switches gaits. To do this, the researchers installed sensor inputs into the periodic patterns of the CPG, allowing the robot to choose the right gait from a specific pattern. It can be pre-programmed with specific patterns, or learn them through experience. For example, the robot can learn how to use as little energy as possible when walking up a slope. When it reaches a hill, its sensors tell it to lower its energy, and it does so by finding a less energy-consuming gait. The next time it attempts to climb a slope, it will automatically know to which gait to switch.

This property makes the Gottingen robot quick and flexible, and further innovations are planned to equip it with a memory device that serves instead of sensors. With this, the robot would be able to maneuver around a terrain even faster, because it will know exactly which gaits to use ahead of time. It would eliminate the need to scan each obstacle, and make it more independent.

[source Max Planck Society for the Advancement of Science.]

Scientific Memory Games: Improve your cognitive abilities.

Go to Source

Posted in AI & Robotics.