Researchers have been inspired by insects to create an anti-collision detector for cars. A much less energy-intensive solution than existing technologies.
Locusts are excellent drivers: they fly in swarms of millions and rarely bump into each other. And this despite simple neural functioning. Few neural resources, for few collisions. If we translate this under the prism of robotics: little energy for few accidents. Why not take inspiration from insects for the anti-collision systems of our future vehicles?
This curious idea is that of researchers from the Department of Engineering and Mechanics at the University of Pennsylvania. They have created a new anti-collision detector for cars, inspired by the reaction of crickets when approaching an obstacle. Their results were published in early 2023 in the journal ACS Nano.
” Insects — like humans — combine several visual information [vitesse, variation de taille ou de forme par exemple] to find out about the danger of an approaching obstacle”, explains Julien Serres, researcher in bio-inspired robotics at the University of Aix-Marseille, who did not participate in the study. This information allows them to assess the time before a potential collision, called the time before impact, and to trigger a flight reaction ” without even having to mentally represent the situation “, he assures.
Objective: to imitate the anti-collision neuron
In locusts, this anti-collision reaction takes place at the level of the LGMD (lobula giant movement detector) neuron. “ This neuron is entirely dedicated to collision detection. By having just one of these neurons in each eye, the cricket detects obstacles and predators while consuming a tiny amount of energy. says University of Pennsylvania researcher Thomas Schranghamer, co-author of the study.
Source : Insect-Inspired, Spike-Based, in-Sensor, and Night-Time Collision
Detector Based on Atomically Thin and Light-Sensitive Memtransistors
As an object approaches, before the insect avoids it, there is a spike in neuronal excitation of the LGMD neuron. The idea of researchers at the University of Pennsylvania is therefore to mimic this excitation peak in an electronic circuit, to evaluate the time before impact.
They simulated the three responses of the neuron:
- Inhibition
- excitement
- The escape
“ Inhibitory and excitatory responses occur when a visual stimulus is presented (such as an approaching object). These responses are combined and lead to an escape response, which depends on the speed of the object and its distance », Details Thomas Schranghamer. To trigger this trio of responses, the researchers added sensors to their device to optically detect any obstacle.
No algorithm, little energy
If the idea of the American team is to reproduce a neural functioning, here we are far from an artificial intelligence. ” There is no program, no software, the calculations are made by electrical impulses. All is hardware”, describes Julien Serres. This has an advantage: the treatment consumes little energy.
Little energy, or at least less than the technologies currently used in vehicles. Because today the anti-collision detectors used in our vehicles are based on pulses of light or radio waves which measure the time of reflection. They therefore require running algorithms to extract the characteristics of images that consume a lot of energy.
By way of comparison, and according to the estimates of Thomas Schranghamer, the consumption of their device is a billion times lower than that of current technologies (of the order of a nanojoule for about twenty seconds of application, against a few joules for d other technology).
Towards bio-inspired robotics for cars
The objective is, in the long term, to integrate this detector into the anti-collision system of your cars. ” This could be used in an anticipatory braking system for example. As we know the inertia of the vehicle, we are able to trigger a coherent braking reaction “says Julien Serres. ” Their adoption will depend on the needs, and may in particular improve existing technologies predicts Thomas Schranghamer. But the question of the cost of this kind of technology arises before a commercial deployment.
“This could be used in an anticipatory braking system”
” What all these little beasts teach us is that there is no need for megapixel cameras or geolocation to move precisely through space. Flies, for example, have very blurred vision – 100 times less fine than human vision – and their avoidance reaction is therefore late but rapid, because adjusted to their size. », illustrates Julien Serres.
Insects direct us towards a different, more sober robotics. Thomas Schranghamer adds: “ If we magnify the line, insects can certainly not do many things, but what they do, they do very well and very efficiently. It’s up to us to glean good ideas from these little animals.
If you liked this article, you will like the following ones: do not miss them by subscribing to Numerama on Google News.
