Researchers Used Locust Brains to Substantially Improve Self-Driving Cars
Detecting potential collisions accurately and responding with a timely escape is crucial in robotics and autonomous vehicle safety
A study published yesterday, 24 August 2020, in the journal Nature Electronics, suggests that locusts have a unique aspect to their vision among insects.
Saptarshi Das, co-author and assistant professor of engineering science and mechanics at Penn State University (PSU) said to EurekAlert "We are always looking for animals with unusual abilities, ones that do something better than humans. Insect vision is something people use regularly to design automatic systems... so we started looking at how it works and, locusts are just incredible. What these creatures can do is very humbling."
Locusts are not considered a good omen, you could take farmers' or the bible's opinion on it. What's interesting is how these insects avoid colliding with each other even when in swarms with numbers reaching 80 million insects.
What makes locusts stand out?
Locusts achieve this feat via a specialized neuron called Lobula Giant Movement Detector (LGMD). Graduate student Darsith Jayachandran explains that the neuron receives two signals and compares them constantly. The first signal detects closeness. When one locust approaches another, its appearance becomes larger and this excites the approached locust's LGMD. The second signal monitors the rotational velocity of the approaching locust relative to the approached locust.
And this is exactly what makes these insects stand out. They have two different means of detecting and reacting to potential collisions. Thanks to their eerie eye shape, locusts have a quite wide field of vision.
So they share the role of supplying the LGMD with the required input, one handles the seeing part while the other calculates the relative rotational velocity. When the LGMD combines these two inputs it triggers an escape response when the stimuli become strong enough.
The first author Darsith Jayachandran explains, "Because the neuron has two branches, the locust computes the changes in these two inputs and realizes that something is going to collide. So the avoiding locust changes direction."
Application to autonomous vehicles
The researchers state that previous work implementing a similar anti-collision measure to self-driving cars has been encouraging for them. But these systems had some major drawbacks, such as their impractical size and high energy consumption. They argue their design is more compact and energy-efficient and could be a breakthrough in this application.
To mimic LGMD's function, the team designed a photoreceptor under 0.001 to 0.005 mm and placed it atop a small flash memory cell. When incoming light increases an internal inhibitory signal decreases.
The team tested the system in a simulated environment. It worked, the car was able to detect collisions before they could happen, but due to the limited depth and rotational perception, the car was unable to decide which way to move to avoid collision.
Now, the researchers are planning to expand their stimuli environment to react to different objects via conditioning the system to different speed, rotation, and light intensity configurations. They hope to develop an applicable and feasible collision avoidance system for autonomous cars and robots.
Akhlesh Lakhtakia, Evan Pugh University Professor, has received a $300,000 grant from the Criminal Investigations and Network Analysis Center to explore a technique for creating 3D holograms of fingerprints.