A synthetic, insect-inspired “brain” can guide a robotic dog using much less energy and weight than conventional approaches require.
The control system, built by UK start-up company Opteran Technologies, mimics how honeybees and other insects navigate. It contains a computer chip and two cameras, which provide a 360-degree view, and can be connected to various robots and drones.
Advertisement
The package weighs only 30 grams and draws less than 3 watts of electricity. That is a fraction of what is required by most robot control systems, says Opteran CEO . “Our algorithms are so lightweight that we can run them on tiny chips, and they run tremendously fast as well,” he says.
The company’s technology builds on the research of at the University of Sheffield, UK, who is also Opteran’s chief scientific officer. Using a combination of behavioural studies and recordings of neural activity, his team tried to reverse-engineer the brain circuits insects use in navigation.
The software that runs on these is similar to the neural networks used by artificial intelligence, says Marshall. But rather than learning how to solve tasks by training on lots of data, it is hardwired to mimic the brain structures responsible for insects’ innate navigational abilities.
The researchers have now shown that a pair of their algorithms can guide a robot dog through a maze of cardboard boxes. The first one stitches together input from the cameras to create a stabilised, panoramic view. The second one mimics how insects analyse optical flow – essentially motion in a visual scene – to estimate velocity and avoid collisions.
Opteran is also working on algorithms that let robots map their environment and make decisions by weighing up different objectives. The company’s technology is being tested by several clients, including for operating delivery drones and inspection robots in mines.
