Engineers Create Low-Power Collision Detector Inspired by Locusts

September 09, 2020 by Luke James

Engineers say they’re creating a low-power collision detector that mimics the way locusts avoid one another when flying in swarms.

It’s not all too often that you see a plague of locusts. And if you were to witness one, chances are that you’d be trying to get away from the swarm. Personal fears and trepidations aside, there is something undeniably fascinating about the way that millions of these insects are able to fly across the sky without colliding with one another. 

Now, researchers at Penn State say that they’re creating a low-power collision detector that is able to mimic the built-in locust avoidance response—known as the lobula giant movement detector (LGMD)—enabled by a specialized collision-avoidance neuron. If successful, the detector could be used to help self-driving cars, drones, and robots avoid collisions.


How Locusts Avoid Collisions

Inspired by how locusts avoid collisions, the Penn State team began looking into the science behind it. 

When a locust is flying as part of a swarm, its LGMD neuron receives two different signals. The first signal received is an image of an approaching locust, which falls on the avoiding locust’s eye. The closer this “invading” locust gets, the stronger the excitation signal becomes. The second signal is the angular velocity of the invader relative to the avoiding locust. 


Penn State researchers are developing a collision detector

Penn State researchers are developing a collision detector for vehicles that mimics an avoidance neuron in locusts that allows them to fly in swarms without colliding. Image used courtesy of News Break

Because the neuron has two branches, the avoiding locust accounts for the changes in these two signals and realizes that a collision is going to occur, so it changes its direction.


A Nanoscale Collision Detector

In a bid to mimic this response, the Penn State researchers report in Nature Electronics article that they developed a nanoscale collision detector by using monolayer molybdenum sulfide as a photodetector. This photodetector was then placed on top of a programmable floating gate memory architecture that’s able to copy the locust’s neuron response by using only a tiny amount of energy. 

The photodetector causes an increase in device current when an oncoming object is detected—the “excitatory signal”—while the programmable memory stack always causes a decrease in current—the “inhibitory signal.” When an object approaches, this excitatory signal is added to the inhibitory stimuli, causing a change in the device current. This mimics the LGMD neuron response that takes place in swarming locusts. 


Low Energy, Fast Response

"While locusts can only avoid collisions with other locusts, our device can detect potential collisions of a variety of objects at varying speeds," said Saptarshi Das, assistant professor of engineering science and mechanics at Penn State.

And just like in locusts, the detector’s response takes place over the course of a few hundred milliseconds. This quick reaction time, in conjunction with low energy use, makes the detector attractive for use as a mechanized collision detector in particularly large, heavy, or potentially dangerous applications such as autonomous vehicles. 


The Penn State researchers claim their collision detector can respond in two seconds

The Penn State researchers claim their collision detector can respond in two seconds. Image used courtesy of Jennifer M. McCann/Penn State

Although the researchers have been successful in mimicking the LGMD neuron response in their detector, plenty of work remains to be done and the technology is a long way off from being commercialized. So far, the device has only been tested with objects on a direct collision path.

Testing still needs to be carried out on responses to other situations.