Autonomous Vehicles Avoiding Collisions Like Locusts
At CES2021 (aka Consumer Electronics Show), Mobileye released their latest Radar and LiDar technology. Using Intel's silicon expertise, Mobileye designed a more compact and inexpensive integrated LiDar solution. This improved technology moves the world in the inevitable direction of autonomous vehicles. Meanwhile, Penn State assistant professor of engineering Saptarshi Das has been inspired by locusts to design a collision avoidance system.
Locusts move in swarms that number thousands but rarely collide with each other. They have evolved with a compound eye providing a wide range of vision. Directly connected to each eye sub-component are two neuron pathways that provide an excitation and impeding voltage, respectively, to a summation cell. This system is called the Lobula Gangula Movement Detector (LGMD). The excitation neuron receives the eye's signal of an approaching object based on the object's change of size, creating an increasing voltage. The second neuron receives the eye's indication of the approaching object's relative velocity, creating a decreasing voltage. This impeding signal eliminates background "noise," such as objects not on a collision course. Within 50ms, the summation cell adds the two neurons' non-linear voltages and provides a direct output signal to the flight control components. Thus, the locust maintains course or changes direction.
Volvo investigated this locust LGMD algorithm for vehicle collision avoidance. However, the energy consumption of the human-inspired design proved to be an issue. As with most of Nature's mechanisms, the Locusts collision avoidance system tends toward energy efficiency and optimization. The challenge is, how can engineers mimic the Locusts collision avoidance sensory with its energy efficiency?
Enter Saptarshi Das, who has discovered a low energy solution to bring this biomimetic solution back to the design table. He mounted a small photoreceptor on a floating-gate transistor. A light signal increases the current in the photoreceptor, and a programmed voltage pulse stimulates the transistor creating a decreasing current. The non-linear summation of the increasing current photoreceptor and the decreasing current transistor creates an output signal to activate an avoidance maneuver.
While initial studies have focused on head-on collisions, future designs will investigate 360-degree collision avoidance needed for next-generation autonomous vehicles. Perhaps LiDar and Radar companies like Mobileye, Bosch, and Luminar could benefit from Nature's intelligent sensing and responding methods.
Jayachandran D, Oberoi A, Sebastian A, Choudhury TH, Shankar B, Redwing JM, Das S. 2020. A low-power biomimetic collision detector based on an in-memory molybdenum disulfide photodetector. Nat Electron. doi:10.1038/s41928-020-00466-9.
